# Lessons from Building LinkedIn's GenAI Platform

Xiaofeng Wang, LinkedIn

AI Engineer Summit 2025

[Talk page](https://ai.engineer/talks/n9rjuBuShko-lessons-from-building-linkedins-genai-platform) · [Video](https://www.youtube.com/watch?v=n9rjuBuShko)

## Lessons from Building LinkedIn's GenAI Platform

LinkedIn’s path from generated articles to recruiting agents shows how product needs shape orchestration, memory, observability, and the team responsible for them.

Familiarity with LLM prompts, APIs, and basic Python will help; retrieval, memory, and agent orchestration are explained as they arise.

### A prompt, an article, and two technology stacks

[1:11](https://www.youtube.com/watch?v=n9rjuBuShko&t=71s)

LinkedIn’s [Collaborative Articles](https://news.linkedin.com/2023/march/collaborative-articles) began with a simple interaction: generate an article, then invite members to contribute their expertise. Xiaofeng Wang describes this 2023 product as LinkedIn’s first formal GenAI feature and identifies **GPT-4** as the model used to produce the long-form content. From the application’s perspective, the mechanism was straightforward: a prompt went in and a string came out.

Supporting that interaction required a gateway to centralize model access and Python notebooks for prompt engineering. But development and serving lived in different worlds: Python on the backend and Java in the online application. These were useful components, not yet what Wang would call a platform. The initial architecture solved access to generation; it did not yet provide a shared way to build richer AI experiences.

![Slide pairs a Collaborative Articles screenshot with six boxes: LLM Playground, Shared Java Midtier, GAI Gateway, Prompt Python Notebooks, Spreadsheet for Eval, and Prompt Resolution Lib.](https://ai.engineer/api/talk-essay-media/5066082f274a5f092c4f952b3d171d6ffcc3e8b418e437a8a80d19561634500f)

Collaborative Articles alongside the early GenAI application platform components.

### Personalization makes shared infrastructure necessary

[2:28](https://www.youtube.com/watch?v=n9rjuBuShko&t=148s)

The limitation became apparent when products needed LinkedIn’s own data. By mid-2023, the next generation of experiences—internally called Copilot or Coach—needed to answer personalized questions. Consider job fit: the application looks at a member’s profile and a job description, then uses **retrieval-augmented generation**, or RAG, to provide recommendations about whether the member fits the role. A generic article prompt cannot supply that context by itself.

The team built a Python SDK on top of [LangChain](https://github.com/langchain-ai/langchain) to orchestrate LLM calls and integrate LinkedIn’s infrastructure. Developers could assemble an application through this shared layer instead of connecting each infrastructure service themselves. This also motivated unifying the technology stack: transferring prompts developed in Python into Java was costly and introduced errors.

Two additional capabilities made the SDK more than a wrapper around model calls. A **prompt source of truth** versioned prompts and provided structure for meta prompts. **Conversational memory** tracked LLM interactions and retrieved content so that the application could inject that context into subsequent interactions. Prompt management controlled the instructions; memory supplied the evolving context needed for a conversational product.

### Recruiting agents need execution and discoverable skills

[4:31](https://www.youtube.com/watch?v=n9rjuBuShko&t=271s)

The next product step was [LinkedIn Hiring Assistant](https://www.linkedin.com/business/talent/blog/talent-acquisition/introducing-hiring-assistant), which Wang describes as the company’s first real multi-agent system. It launched in charter in 2024, initially for selected customers and LinkedIn recruiters. Wang frames the recruiting workload broadly: posting jobs, evaluating hundreds of candidates, and reaching out to them. The system’s purpose was to automate tedious work across that process, rather than merely return advice in a conversation.

That change expanded the Python SDK into a distributed agent orchestration layer. It had to handle distributed execution, retry logic, and traffic shifting. An agent that chooses and performs actions needs more operational support than an application that makes a single model request.

Actions also require a usable interface to enterprise APIs. The team built tools for developers to publish their APIs into a centralized **skill registry**. The registry addressed both discovery—finding an available skill—and invocation—calling it from an application. Publishing an API as a skill made the existing service accessible through the agent platform’s common machinery.

### Remember interactions and inspect autonomous behavior

[6:11](https://www.youtube.com/watch?v=n9rjuBuShko&t=371s)

Memory evolved alongside execution. Beyond retaining conversational content, **experiential memory** extracted, analyzed, and inferred contextual knowledge from interactions between users and agents. The platform organized memory into working, long-term, and collective layers. The distinction is useful: preserving an interaction gives the system a record; extracting contextual knowledge makes that interaction useful as experience.

Autonomy made observability essential. Because an agent can decide which API or LLM to call, its behavior is difficult to predict in advance. LinkedIn built an in-house solution on [OpenTelemetry](https://opentelemetry.io/docs/) to capture fine-grained telemetry that could be used to replay agent calls. An analytics layer on top of those records then helped guide future optimization. The platform needed visibility into the agent’s execution, not just its final response.

### One interface to a larger AI ecosystem

[7:37](https://www.youtube.com/watch?v=n9rjuBuShko&t=457s)

These investments formed four main areas of the GenAI application platform:

| Area | Responsibility |
| --- | --- |
| Orchestration | Coordinate model and agent execution |
| Prompt engineering | Manage instructions and their versions |
| Tools and skills | Support discovery and invocation |
| Content and memory | Supply context across interactions |

The application platform brought these concerns together, but it did not own the entire AI ecosystem.

Sister teams handled model work such as fine-tuning open-source models, responsible AI policies and standards, and the machine-learning infrastructure that hosted models. The GenAI platform’s role was to give application developers a **unified interface** to that ecosystem. Developers could use those capabilities without first learning every underlying service and integration.

Wang gives model selection as a concrete example: in LinkedIn’s SDK, changing one parameter in one line of code could switch an application from an OpenAI model to an on-premises model. Developers still had to do prompt engineering for the selected model; the abstraction removed infrastructure integration work, not differences in model behavior. Centralization also created a place to enforce best practices and governance, helping developers build applications efficiently and responsibly.

![Layered diagram covers orchestration, prompt engineering, tools, models, responsible AI, AI platform, and context and memory. Five bullets describe integration, developer experience, reliable development, and responsible AI practices.](https://ai.engineer/api/talk-essay-media/4a726497474f6f92d1b386dedfb5a3739aba7d2942a923ac2b1139961a485dcf)

LinkedIn’s GenAI ecosystem and the platform’s integration and governance responsibilities.

### The boundary between optimization and serving moves

[9:28](https://www.youtube.com/watch?v=n9rjuBuShko&t=568s)

The platform emerged piece by piece, which raises a reasonable question: with so many vendors offering GenAI tooling, does a company need this platform capability, and should it build or buy it? Wang’s answer addresses the need for the capability rather than selecting a vendor or prescribing that every component be built internally. The reason is a change in how AI applications are developed.

In the traditional arrangement Wang describes, model optimization and serving have a relatively clear boundary. AI engineers and product engineers can use different stacks and need not work in the same codebase. In a GenAI application, that separation fades: engineers on either side can change the overall system’s performance.

| Development concern | Traditional model-centered system | GenAI application |
| --- | --- | --- |
| Optimization and serving | Separate phases | Intertwined work |
| Engineering stacks | Can remain separate | Need shared tooling |
| System improvement | Divided by role | Spans both groups |

Once optimization reaches into application behavior, tooling and engineering practices have to support both groups working on the system.

Wang connects this to Berkeley AI Research’s concept of a **compound AI system**: a system that tackles AI tasks through interacting components, including multiple model calls, retrievers, or external tools. Improving such a system requires skills that cross the traditional AI-engineering and product-engineering boundary. A shared platform bridges that gap by making the components usable through common interfaces and practices.

### Build the team’s range through collaboration

[11:34](https://www.youtube.com/watch?v=n9rjuBuShko&t=694s)

That platform remit creates a demanding hiring profile. Looking at the strongest engineers on his team, Wang identifies infrastructure integration skills, a developer product-management sensibility for interface design, AI and data-science knowledge, and the ability to learn new techniques while remaining hands-on. Finding all of these in one person is rare—rare enough that he jokes the candidate would be worth more than a unicorn. Hiring therefore requires deliberate trade-offs.

- **Core engineering skills:** Prioritize strong software engineering over existing AI expertise.
- **Potential:** Hire for the ability to learn rather than relying mainly on experience or degrees, because the field changes quickly.
- **Complementary backgrounds:** Build the missing range across the team. LinkedIn’s mix included full-stack engineers, data scientists, AI engineers, data engineers, recent research-university graduates, and people with startup experience.

The development mechanism was joint project work. Wang reports that strong engineers acquired complementary skills through collaboration and grew toward the broader ideal profile.

![Dream Team slide shows an ideal engineer profile beside four bullets: Focus on Core Skills, Hire for Potential, Promote Diversity, and Foster Collaboration.](https://ai.engineer/api/talk-essay-media/76ed03072a8e39998af76a76a90a5130fb5d401aeaded0605628346b278821bb)

Building a team through core engineering skills, potential, diversity, and collaboration.

Critical thinking matters after the team is hired, too. Wang describes a recurring team premise: what they are building may become outdated within a year, or even in less than six months. That premise drives an ongoing practice of evaluating open-source packages, talking with vendors, and proactively deprecating internal solutions. Building the platform does not require defending every component indefinitely.

### Choose Python and treat prompts as production parameters

[14:30](https://www.youtube.com/watch?v=n9rjuBuShko&t=870s)

Returning to the technology stack, Wang recommends Python where possible. LinkedIn began with Java and Python, debated the choice internally, and ultimately chose Python because of its alignment with research and the open-source ecosystem. Wang also reports that Python proved scalable in LinkedIn’s experience; he supplies no workload measurements for a language comparison.

**Prompts deserve the operational discipline applied to model parameters.** A prompt source of truth needs robust version control because changing instructions changes application behavior. An accidental production edit can create unwanted side effects even when the surrounding application code remains unchanged.

A small Python example makes the distinction between authoring a prompt and selecting a production version concrete. Here, the application explicitly selects `job-fit-v1`; adding `job-fit-v2` to the catalog does not change that selection:

```python
from types import MappingProxyType

PROMPTS = MappingProxyType({
    "job-fit-v1": (
        "Assess this member's fit for the job using the supplied facts.\n"
        "Profile: {profile}\nJob description: {job_description}"
    ),
    "job-fit-v2": (
        "Compare the supplied profile with the job requirements.\n"
        "Separate supported matches from missing information.\n"
        "Profile: {profile}\nJob description: {job_description}"
    ),
})

PRODUCTION_PROMPT_VERSION = "job-fit-v1"


def render_job_fit_prompt(profile: str, job_description: str) -> str:
    template = PROMPTS[PRODUCTION_PROMPT_VERSION]
    return template.format(
        profile=profile,
        job_description=job_description,
    )
```

This illustrates explicit version selection, not LinkedIn’s internal SDK. The production safeguard depends on preserving released versions and making changes to the selected version deliberate.

The other essential investments follow the product’s needs. Memory injects rich enterprise data into agent experiences. Tooling around existing APIs turns them into skills that agents can call easily. Together, these components give applications controlled instructions, relevant context, and access to actions.

### Grow from immediate needs into developers’ workflows

[16:02](https://www.youtube.com/watch?v=n9rjuBuShko&t=962s)

The route to adoption starts smaller than the finished architecture suggests. Solve an immediate need instead of building a full platform at the outset. LinkedIn began with a simple Python orchestration library and expanded into the broader set of components as product requirements grew. The useful unit of progress was a capability developers could use, not completion of the entire platform diagram.

Scaling also meant reusing infrastructure. LinkedIn used its existing messaging infrastructure as a memory layer, an approach Wang describes as cost efficient and scalable. The lesson is to look for infrastructure that already provides the foundations a new capability needs, rather than assuming that an AI-specific requirement demands an entirely new storage system.

Developer experience determines whether those capabilities are adopted. The platform exists to make developers productive, so its technology needs to fit their existing workflows. A technically capable system that makes everyday development harder misses the reason for building it.

![Full-screen slide with three blue columns labeled Start Simple, Evolve Gradually; Focus on Scalable Solutions; and Prioritize Developer Experience.](https://ai.engineer/api/talk-essay-media/e05c26e087ffe649884dc5d8d7f8394c50bcb98a6dca2b3ddc7b4bd3f31d4278)

Platform adoption lessons: start simple, build for scale, and prioritize developer experience.

For the lower-level implementation details, Wang points to the team’s [engineering article on LinkedIn’s GenAI application stack](https://www.linkedin.com/blog/engineering/generative-ai/behind-the-platform-the-journey-to-create-the-linkedin-genai-application-tech-stack), by Karthik R., Xiaofeng Wang, and Sandeep Jha.

## Resources

- [Behind the platform: LinkedIn’s GenAI application tech stack](https://www.linkedin.com/blog/engineering/generative-ai/behind-the-platform-the-journey-to-create-the-linkedin-genai-application-tech-stack) — The team's technical account of Python adoption, LangChain integration, prompt management, skill discovery, memory infrastructure and model access.
- [LinkedIn Hiring Assistant launch announcement](https://www.linkedin.com/business/talent/blog/talent-acquisition/introducing-hiring-assistant) — The October 2024 charter launch announcement describes recruiting tasks, human oversight and planned capabilities.
- [Introducing LinkedIn collaborative articles](https://news.linkedin.com/2023/march/collaborative-articles) — LinkedIn explains how AI-assisted article starters connect with contributions from relevant member experts.
- [LangChain](https://github.com/langchain-ai/langchain) — The public framework repository underlying LinkedIn's historical Python SDK; its current code has evolved since the talk.
- [OpenTelemetry documentation](https://opentelemetry.io/docs/) — Current documentation for the telemetry project used as the foundation of LinkedIn's in-house observability solution.

## Slides

### GenAI Product: Collaborative Articles

LinkedIn launched Collaborative Articles in 2023 as a prompt-in, string-out GenAI application. The speaker uses it to introduce the product experience the platform supports.

[1:14](https://www.youtube.com/watch?v=n9rjuBuShko&t=74s) [Slide introducing Collaborative Articles with a LinkedIn article example.](https://assets.ai.engineer/sites/aiecode2025/sha256/27c2075ddabf52c8e154be654b04f0b34fc19564b025b81b62ef0c9992b71422.webp)

GenAI App Platform Journey: Exploring the Unknown

GenAI Product

Assess the situation

Before you take any action, you need to assess the situation and understand the root cause of the difficulty. Is the team member struggling with a personal or professional issue, such as stress, burnout, or lack of skills? Are they frustrated with the team’s goals, processes, or communication? Are they unaware of the impact of their behavior on others, or do they have a different personality or work style? By asking yourself these questions, you can gain a clearer picture of the problem and its possible solutions.

Ayaka Kobayashi

In my role, I’ve learned the value of understanding tough team situations. Take a recent example: a senior chef was frustrated with the team’s performance. I didn’t rush into solutions. Instead, we had a conversation and found that their concerns were work-related, not personal. We provided training and shifted responsibilities. This experience reinforced the importance of digging into the root causes. It led to a more collaborative and productive work environment, which aligns with Marriott International’s high standards.

Ellis Morgan

I’ve found that managing challenging team members involves careful assessment. It’s like analyzing complex financial data — you need to understand the underlying issues. Recently, one team member consistently missed deadlines. Instead of reprimanding, I talked to them and discovered they were overwhelmed. We reassigned tasks and provided support. This approach helped

Collaborative Articles

Interface text:
LinkedIn
Search
Add your perspective
Follow
Like
Share
AI Engineer Summit NYC 2025
LinkedIn
3

### GenAI App Platform Journey: Exploring the Unknown

Collaborative Articles relied on early tools including a gateway that centralized model access and Python notebooks for prompts. Java served the online experience; Python ran the backend. The speaker did not yet call this a platform.

[1:57](https://www.youtube.com/watch?v=n9rjuBuShko&t=117s) [Slide pairs Collaborative Articles with platform components including a playground, gateway, notebooks, evaluation spreadsheet, Java midtier and prompt resolution library.](https://assets.ai.engineer/sites/aiecode2025/sha256/0d1f183d3ac9d560524dae1fe6f26caa372dcdc554403051202c224fc774eb67.webp)

GenAI App Platform Journey: Exploring the Unknown

GenAI Product
Collaborative Articles

GenAI App “Platform”
LLM Playground
Shared Java Midtier
GAI Gateway
Prompt Python Notebooks
Spreadsheet for Eval
Prompt Resolution Lib

Interface text:
Assess the situation
[illegible]
Add your perspective
Ayaka Kobayashi
[illegible]
Ellis Morgan
[illegible]
Like
Share
LinkedIn
3
AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER

[2:12](https://www.youtube.com/watch?v=n9rjuBuShko&t=132s) [A collaborative article example alongside early platform tools, including a gateway, notebooks, and an evaluation spreadsheet.](https://assets.ai.engineer/sites/aiecode2025/sha256/28a4dd7967bcc33db197c1326398436a4edd5ef84873f29e019eb88cead8a004.webp)

GenAI App Platform Journey: Exploring the Unknown

GenAI Product
Collaborative Articles

GenAI App “Platform”
LLM Playground
Shared Java Midtier
GAI Gateway
Prompt Python Notebooks
Spreadsheet for Eval
Prompt Resolution Lib

Interface text:
How do you manage difficult team members?
[illegible]
Top experts in this article
[illegible]
Earn a community badge
[illegible]
Assess the situation
[illegible]
Add your perspective
Like
Share
LinkedIn
3
AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER

### Building the Python Orchestration Foundations

LinkedIn’s job seeker assistant uses a RAG process with profiles and job descriptions to offer personalized job-fit recommendations. The second generation began in mid-2023 to bring richer data into the experience.

[2:32](https://www.youtube.com/watch?v=n9rjuBuShko&t=152s) [Slide introducing a job seeker experience with a mobile job listing.](https://assets.ai.engineer/sites/aiecode2025/sha256/cf59ef192019d5a90ec8650a960612caab893b7e793542dbc33f007ae4665ff7.webp)

GenAI App Platform Journey: Building the Python Orchestration Foundations

GenAI Product

New Job Seeker Experience

Interface text:
Customer Service Manager
Flexis
Phoenix, AZ
2 days ago · 30 applicants
$75K-110K/yr (est.) · Full time
501-1000 employees · IT industry
David Yuan and 6 connections can refer you
PREMIUM
Get advice on the right jobs for you
Am I a good fit for this job?
[illegible]
Apply
Save
Meet the hiring team
[illegible]
About this job
[illegible]

AI Engineer Summit NYC 2025
LinkedIn
4

### Job Seeker Assistant Response

The demo answers “Who works at Flexis?” with people to contact and follow-up prompts. The speaker says the experience also uses profiles, job descriptions, and RAG to recommend whether a job is a good fit.

[2:39](https://www.youtube.com/watch?v=n9rjuBuShko&t=159s) [A mobile job seeker assistant displays a response about people at an employer.](https://assets.ai.engineer/sites/aiecode2025/sha256/3c96de466911d5a3f11c385f48baa0cd703b1b7552ecdc168c1309b6ca622538.webp)

GenAI App Platform Journey: Building the Python Orchestration Foundations

GenAI Product

[illegible]w Job Seeker Experience

Interface text:
Who works at Flexis?

Here are some people in LinkedIn’s professional network that hold a Customer Service position at Flexis. Reach out to connect and learn more about their experience.

[illegible]
[illegible]

How can I send a personalized message?
What is the work and life balance like at Flexis?
Ask a question...

AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER
LinkedIn
4

### GenAI App Platform Journey: Python Orchestration Foundations

LinkedIn’s job seeker experience uses a profile and job description in a RAG process to recommend whether someone is a good fit for a job.

[2:57](https://www.youtube.com/watch?v=n9rjuBuShko&t=177s) [The product-only build shows a LinkedIn job seeker experience on a phone.](https://assets.ai.engineer/sites/aiecode2025/sha256/7d4b65f0d8b872348f3949bf76f78d746c7ac58729cf1957a6918ee36463dd16.webp)

GenAI App Platform Journey: Building the Python Orchestration Foundations

GenAI Product

[illegible]w Job Seeker Experience

Interface text:
Customer Service Manager
Flexis
Phoenix, AZ
2 days ago · 30 applicants
[illegible]
Am I a good fit for this job?
[illegible]
Apply
Save
Meet the hiring team
[illegible]
About this job
[illegible]

AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER
LinkedIn
4

### GenAI App Platform Journey: Python Orchestration Foundations

LinkedIn built a LangChain-based Python SDK to orchestrate LLM calls and integrate its infrastructure, alongside app hosting, prompt versioning, and conversational memory for the job seeker experience.

[3:13](https://www.youtube.com/watch?v=n9rjuBuShko&t=193s) [A job seeker experience is paired with four platform components: app framework, app hosting, prompt source of truth, and conversational memory.](https://assets.ai.engineer/sites/aiecode2025/sha256/7e0274e9a3065dc6aa28486f3150e75b1f04be9a7ace78ca81201377dbc0d344.webp)

GenAI App Platform Journey: Building the Python Orchestration Foundations

GenAI Product
New Job Seeker Experience

GenAI App Platform

App Framework
Thin LangChain wrapper integrating LinkedIn’s infrastructure, offered as a versioned Python library.

App Hosting
Python gRPC service easily deployable across all LinkedIn fabrics.

Prompt SOT
A submodule in the app framework using Jinja to provide more structure around modularization & versioning.

Conversational Memory
Infrastructure to store LLM interactions, retrieve context, and inject it into prompts for coherent product experiences.

AI Engineer Summit NYC 2025
LinkedIn
4

Interface text:
Who works at Flexis?
Here are some people in LinkedIn’s professional network that hold a Customer Service position at Flexis. Reach out to connect and learn more about their experience.
[illegible]
How can I send a personalized message?
What is the work and life balance like at Flexis?
Ask a question...

### GenAI App Platform Journey: Agentic Era

LinkedIn Hiring Assistant uses multiple agents to help recruiters evaluate candidates and reach out. The screens show a candidate list and a detailed assessment of one candidate.

[4:31](https://www.youtube.com/watch?v=n9rjuBuShko&t=271s) [A LinkedIn recruiting interface illustrates the hiring agent.](https://assets.ai.engineer/sites/aiecode2025/sha256/25bf946376ce71f24dcc2f0d4fe3861cdb384167062ae4528b5d7308010579c6.webp)

GenAI App Platform Journey: Agentic Era

GenAI Product

LinkedIn Hiring Agent

AI Engineer Summit NYC 2025
LinkedIn
5

Interface text:
Recruiter
Projects  Jobs  Campaigns  Reports
Growth Marketer  Pipeline
I've sourced candidates who are a good fit for this role.
Start outreach
All candidates  32
Archived candidates  10
Active stages
Not contacted  22
Contacted  0
Replied  0
Added by: Hiring Assistant
Clear all
Antoine Martel  Sourced
Antoine is a top candidate
[illegible]
Change stage
Archive
Amala Jain  Sourced

[4:34](https://www.youtube.com/watch?v=n9rjuBuShko&t=274s) [Product-only slide showing the LinkedIn Hiring Agent candidate interface.](https://assets.ai.engineer/sites/aiecode2025/sha256/23822ee77cfd016a3492a5fb861133bc29f989e720e47e3ba00bba09805fd9f0.webp)

GenAI App Platform Journey: Agentic Era

GenAI Product

LinkedIn Hiring Agent

AI Engineer Summit NYC 2025
LinkedIn
5

Interface text:
From search results
Help me improve
Antoine Martel  Sourced
Senior Growth Marketing Manager @ Flexis
Add email
Add phone number
Public profile
Change stage
Archive
Profile  Projects (6)  Messages (0)  More
Most recent activities
Hiring Assistant added candidates to Growth Marketer
Viewed by Jean Tucker
Attachment (Resume) has been added
Summary
Experienced digital marketer with a passion for growth marketing and launching strategies for tech.
Top skills: Digital marketing, Analytics, CRO, Project Management
Antoine is a top candidate
[illegible]
Required qualifications 3/3
Experience in launching products from 0-1
Digital strategy
Proficiency in data analysis and the use of analytics tools like Google Analytics or Mixpanel
Preferred qualifications [illegible]
An MBA or Master's degree in Marketing or a related field
Over 5 years of experience in growth marketing or a related role, preferably in a startup
[illegible]

### GenAI App Platform Journey: Agentic Era

LinkedIn’s platform orchestrates distributed agents, handling execution, retries, and traffic shifts. A central skill registry lets developers publish APIs for agents to use.

[5:04](https://www.youtube.com/watch?v=n9rjuBuShko&t=304s) [Slide pairs the hiring product with agent framework, API and skills, memory, and observability components.](https://assets.ai.engineer/sites/aiecode2025/sha256/b8574c5a50f243a0ce5bc15793af3f4a59f0211e587804f50676f86a3931e4db.webp)

GenAI App Platform Journey: Agentic Era

GenAI Product
LinkedIn Hiring Agent

GenAI App Platform

Agent Framework
LangChain/LangGraph based simple agents plus in-house solution for the complex distributed agents.

API/Skills
Tools & infra for functional calling to APIs, featuring a centralized skill registry, plugin registration, & dynamic tool retrieval / invocation.

Experiential Memory
Service to extract, infer, & evolve contextual knowledge, including working, long-term, & collective memories.

Observability
Monitoring service built on OTel, providing insights like execution paths, aggregation, & evaluation.

LinkedIn
5

Interface text:
Recruiter · Projects · Jobs · Campaigns · Reports
Good morning, Allison
22 candidates are ready for review and outreach.
Start outreach · Review candidates
Projects with high engagement (7)
Continue recent searches
[illegible]

AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER

### GenAI App Platform Journey: Agentic Era

LinkedIn pairs its Hiring Agent with a platform for distributed agent orchestration, API skills, memory, and observability. The framework extends its Python SDK to larger-scale orchestration.

[5:10](https://www.youtube.com/watch?v=n9rjuBuShko&t=310s) [Expanded platform slide alongside a hiring candidate interface.](https://assets.ai.engineer/sites/aiecode2025/sha256/3d5b639f865671550b63d92b405ab5e75c8921d9c9d784491f58e539a92a8d2c.webp)

GenAI App Platform Journey: Agentic Era

GenAI Product
LinkedIn Hiring Agent

GenAI App Platform

Agent Framework
LangChain/LangGraph based simple agents plus in-house solution for the complex distributed agents.

API/Skills
Tools & infra for functional calling to APIs, featuring a centralized skill registry, plugin registration, & dynamic tool retrieval / invocation.

Experiential Memory
Service to extract, infer, & evolve contextual knowledge, including working, long-term, & collective memories.

Observability
Monitoring service built on OTel, providing insights like execution paths, aggregation, & evaluation.

LinkedIn
5

Interface text:
AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER

Hiring interface: From search results; Help me improve; Antoine Martel; Antoine is a top candidate; Most recent activities; Summary; Recruiting tools; Similar profiles. [illegible]

### GenAI App Platform Journey: Agentic Era

LinkedIn pairs its Hiring Agent with a platform for agent orchestration, API and skill calls, memory, and observability. The distributed orchestration layer handles execution, retries, and traffic shifts.

[5:17](https://www.youtube.com/watch?v=n9rjuBuShko&t=317s) [Expanded platform slide alongside the hiring product home screen.](https://assets.ai.engineer/sites/aiecode2025/sha256/1423c9d005d3a93c22bef25589a38d29b486c7a1a08ea6a5286bdd857278880b.webp)

GenAI App Platform Journey: Agentic Era

GenAI Product
LinkedIn Hiring Agent

GenAI App Platform

Agent Framework
LangChain/LangGraph based simple agents plus in-house solution for the complex distributed agents.

API/Skills
Tools & infra for functional calling to APIs, featuring a centralized skill registry, plugin registration, & dynamic tool retrieval / invocation.

Experiential Memory
Service to extract, infer, & evolve contextual knowledge, including working, long-term, & collective memories.

Observability
Monitoring service built on OTel, providing insights like execution paths, aggregation, & evaluation.

LinkedIn
5

Interface text:
Recruiter
Good morning, Allison
22 candidates are ready for review and outreach.
Start outreach
Review candidates
[illegible]

AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER

[5:20](https://www.youtube.com/watch?v=n9rjuBuShko&t=320s) [Slide showing a hiring agent alongside platform capabilities for agent frameworks, APIs, memory, and observability.](https://assets.ai.engineer/sites/aiecode2025/sha256/555d2bb1801831320e9457dc18599cc978c82bc98eba35a9abdb94d9db40ba10.webp)

GenAI App Platform Journey: Agentic Era

GenAI Product
LinkedIn Hiring Agent

GenAI App Platform

Agent Framework
LangChain/LangGraph based simple agents plus in-house solution for the complex distributed agents.

API/Skills
Tools & infra for functional calling to APIs, featuring a centralized skill registry, plugin registration, & dynamic tool retrieval / invocation.

Experiential Memory
Service to extract, infer, & evolve contextual knowledge, including working, long-term, & collective memories.

Observability
Monitoring service built on OTel, providing insights like execution paths, aggregation, & evaluation.

LinkedIn
5

Interface text:
Recruiter
Growth Marketer
Pipeline
I've sourced candidates who are a good fit for this role.
Start outreach
All candidates
Archived candidates
Active stages
Not contacted
Contacted
Replied
[illegible]

AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER

### GenAI App Platform Journey: Agentic Era

LinkedIn pairs its Hiring Agent with frameworks, API skills, memory and observability. Its OTel-based monitoring tracks execution paths; the speaker says telemetry can replay agent calls.

[6:54](https://www.youtube.com/watch?v=n9rjuBuShko&t=414s) [Full slide showing a hiring agent and four agent-platform capabilities: framework, APIs, memory, and observability.](https://assets.ai.engineer/sites/aiecode2025/sha256/9d6e8ca6bb11196dd4084429ae65b66340ce4ebace93aeeb5a1b7d2d1a874dfc.webp)

GenAI App Platform Journey: Agentic Era

GenAI Product
LinkedIn Hiring Agent

GenAI App Platform

Agent Framework
LangChain/LangGraph based simple agents plus in-house solution for the complex distributed agents.

API/Skills
Tools & infra for functional calling to APIs, featuring a centralized skill registry, plugin registration, & dynamic tool retrieval / invocation.

Experiential Memory
Service to extract, infer, & evolve contextual knowledge, including working, long-term, & collective memories.

Observability
Monitoring service built on OTel, providing insights like execution paths, aggregation, & evaluation.

AI Engineer Summit NYC 2025
LinkedIn
5

Interface text:
Recruiter
From search results
Help me improve
Antoine Martel
Antoine is a top candidate
[illegible]
Most recent activities
Summary
[illegible]

### GenAI App Platform: Orchestration, Tooling & Seamless Integration…

LinkedIn places LangChain/LangGraph in orchestration, prompt storage and versioning in prompt engineering, agents and APIs in tools, and conversational memory, experiential memory, and RAG in context and memory.

[7:34](https://www.youtube.com/watch?v=n9rjuBuShko&t=454s) [Diagram of platform layers for orchestration, prompt engineering, tools, and context and memory.](https://assets.ai.engineer/sites/aiecode2025/sha256/9f7530fc918c3ec49b4de792fca11c3c50ff17bfd24f81e649617bed1bb4e51e.webp)

GenAI App Platform: Orchestration, Tooling & Seamless Integration for Developer Efficiency

Orchestration: LangChain/LangGraph; App Hosting; Observability
Prompt Engineering: Prompt Store; Prompt Versioning
Tools: API / Skills; Agents
Context & Memory: Conversational Memory; Experiential Memory; RAG

Interface text:
AIE
LinkedIn
6
AI LEADERSHIP
HTTPS://AI.ENGINEER

### GenAI App Platform and LinkedIn’s GenAI Ecosystem

LinkedIn’s platform gives developers a unified interface to its GenAI ecosystem. In its SDK, changing one parameter switches between an OpenAI model and an on-prem model.

[8:30](https://www.youtube.com/watch?v=n9rjuBuShko&t=510s) [LinkedIn’s GenAI ecosystem diagram alongside four points on platform focus, integration, developer experience, and application development.](https://assets.ai.engineer/sites/aiecode2025/sha256/0ec7501d667e6354b6fa37d08d0e63c649d6776b8ecdc201c4e3fe726ca3fc79.webp)

GenAI App Platform: Orchestration, Tooling & Seamless Integration for Developer Efficiency

• Focused on orchestration, tools/skills, and prompt management.
• Seamlessly integrates with underlying services in the ecosystem.
• Optimized for the best developer experience.
• Enables fast, reliable application development.

Orchestration: LangChain/LangGraph; App Hosting; Observability
Prompt Engineering: Prompt Store; Prompt Versioning
Tools: API / Skills; Agents
Model: Proprietary Models; Domain Adapted Foundation Models
Responsible AI: Alignment; Red Teaming; Moderation
AI Platform: Training; Serving; Evaluation
Context & Memory: Conversational Memory; Experiential Memory; RAG

LinkedIn’s GenAI Ecosystem

LinkedIn
6

AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER

### GenAI App Platform and Developer Efficiency

LinkedIn’s platform brings orchestration, tools and prompt management together with model, safety and memory services. Its centralized design also gives teams a place to enforce responsible AI practices.

[9:04](https://www.youtube.com/watch?v=n9rjuBuShko&t=544s) [LinkedIn’s GenAI ecosystem layers orchestration, prompt engineering, tools, models, responsible AI, platform services, and memory alongside developer-platform benefits.](https://assets.ai.engineer/sites/aiecode2025/sha256/6e8ed5d97a227670d32edbfd1f2d1fe20e557b6b8bd0b423e33296ad36353754.webp)

GenAI App Platform: Orchestration, Tooling & Seamless Integration for Developer Efficiency

• Focused on orchestration, tools/skills, and prompt management.
• Seamlessly integrates with underlying services in the ecosystem.
• Optimized for the best developer experience.
• Enables fast, reliable application development.
• Enforces best practices, particularly for Responsible AI (RAI) and trust compliance.

LinkedIn’s GenAI Ecosystem
Orchestration: LangChain/LangGraph; App Hosting; Observability
Prompt Engineering: Prompt Store; Prompt Versioning
Tools: API / Skills; Agents
Model: Proprietary Models; Domain Adapted Foundation Models
Responsible AI: Alignment; Red Teaming; Moderation
AI Platform: Training; Serving; Evaluation
Context & Memory: Conversational Memory; Experiential Memory; RAG

LinkedIn
6

Interface text:
AIE
AI LEADERSHIP
HTTPS://AI.ENGINEER

### The Need for a GenAI App Platform

Traditional AI hands a trained model from optimization to serving. In GenAI, models and prompts bring AI and product engineers into the same system optimization and serving phases.

[10:25](https://www.youtube.com/watch?v=n9rjuBuShko&t=625s) [Traditional AI separates model optimization from serving; the GenAI diagram combines models and prompts in system optimization and serving.](https://assets.ai.engineer/sites/aiecode2025/sha256/18b0e72cb26c947857beda3c90f8165a097a6af113e7f8c46a069c92ed5f561c.webp)

The Need for a GenAI App Platform: Everyone is an AI Engineer in GenAI Era

Traditional AI Systems
AI Engineers
Trained Model Artifact
Product Engineers
AI Model Optimization Phase
AI Model Serving Phase

GenAI Systems
AI Engineers
LLMs
Fine-Tuned Models
Prompt
Product Engineers
AI Engineers
AI System Optimization & Serving Phases

AI Engineer Summit NYC 2025
LinkedIn
7

### GenAI app platform and compound AI system optimization

The proposed platform connects LLMs to GenAI products through system optimization, aligning engineers and involving non-technical roles in evaluation while addressing integration and testing skill gaps.

[11:27](https://www.youtube.com/watch?v=n9rjuBuShko&t=687s) [A diagram links LLMs to GenAI products through compound AI system optimization. Bullets describe collaboration, evaluation, responsible AI, and bridging engineering skill gaps.](https://assets.ai.engineer/sites/aiecode2025/sha256/8e112ecde471b73dde942cb02d7b1ed17bb041172c8488558996909e47f40ec4.webp)

GenAI App Platform is Critical to Bridge the New Gaps in Compound AI System Optimization

LLMs → Compound AI System Optimization → GenAI Products

• Align AI and product engineers on prompt optimization efforts.
• Enable non-technical roles in GenAI evaluation and optimization.
• Enforce Responsible AI and trust seamlessly.
• Bridge skill gaps by simplifying API integration for AI engineers and dynamic AI system testing & optimization for backend engineers.

“Compound AI System as a system that tackles AI tasks using multiple interacting components, including multiple calls to models, retrievers, or external tools.”
- Berkeley AI Research (BAIR)

Interface text:
AIE
LinkedIn
8
AI LEADERSHIP
HTTPS://AI.ENGINEER

### Building a Dream Team

LinkedIn prioritizes software engineering skills and hires for potential as experience can become outdated. Its ideal candidate spans infrastructure, interfaces, AI, and data science.

[12:59](https://www.youtube.com/watch?v=n9rjuBuShko&t=779s) [Candidate qualifications alongside recommendations to prioritize core engineering skills and hire for potential.](https://assets.ai.engineer/sites/aiecode2025/sha256/f16639c1baaf02838c9bca7c7dfe42af5dfee9909eb9e5cb42d58675e0348681.webp)

Building a Dream Team: Engineering Skills, Diversity, and Critical Thinking

An ideal candidate is a strong software engineer skilled in infrastructure integration, developer interface design, with a background in AI and data science, the ability to quickly learn new technologies, implement solutions efficiently, and adapt to evolving trends.

• Focus on Core Skills: Prioritize strong software engineering skills over specialized AI expertise.
• Hire for Potential: Seek high-potential individuals with a growth mindset, adaptability, passion for innovation, and a strong sense of ownership.

AI Engineer Summit NYC 2025
LinkedIn
9

## Full transcript

[0:00](https://www.youtube.com/watch?v=n9rjuBuShko&t=0s) [on-hold music] Uh, it's my pleasure here to share our journey on building out LinkedIn's GenAI platform.

[0:23](https://www.youtube.com/watch?v=n9rjuBuShko&t=23s) Uh, my name is Xiaofeng Wang, a manager of GenAI Foundation.

[0:29](https://www.youtube.com/watch?v=n9rjuBuShko&t=29s) Uh, let me try it one more time. Cool. Uh, in today's talk, I'd like to first share our journey on building out this platform, especially on why we're building it, how we build it, and what we're building it.

[0:44](https://www.youtube.com/watch?v=n9rjuBuShko&t=44s) After that, we will talk about, uh, some thought process on why this platform is critical for today's agent world. Um, hopefully after that, you agree with me this is critical component in your component-- uh, uh, in your company, and you also want to build this team.

[1:01](https://www.youtube.com/watch?v=n9rjuBuShko&t=61s) I want to share some tips on how to build such a team, how to hire for such a team. Towards the end, we will share some key takeaways and lessons learned.

[1:11](https://www.youtube.com/watch?v=n9rjuBuShko&t=71s) Before we dive into this application, uh, platform journey, I think it's important to first talk about the GenAI product experience because that's essentially what our platform is supporting for.

[1:26](https://www.youtube.com/watch?v=n9rjuBuShko&t=86s) Back in twenty-twenty three, LinkedIn launched the first, uh, formal GenAI feature called Collaborative Articles. This is a kind straightforward, uh, GenAI feature if we are thinking in today's standard because it's a very simple prompt in, string out type of application.

[1:46](https://www.youtube.com/watch?v=n9rjuBuShko&t=106s) Uh, we leverage ChatGPT-- Uh, I mean, we leverage GPT-4 model, uh, to create the long content, uh, articles on the platform and then invite our members to comment on it.

[1:59](https://www.youtube.com/watch?v=n9rjuBuShko&t=119s) At this stage, our team helped to build some key component behind the scene, including the gateway to centralize the access to the model, uh, some Python notebook for the prompt engineering.

[2:11](https://www.youtube.com/watch?v=n9rjuBuShko&t=131s) Uh, but at this time, we actually have, uh, two different tech stacks, uh, to serve the, uh, experience. In the online phase, we use Java, and in the back end, we use Python.

[2:23](https://www.youtube.com/watch?v=n9rjuBuShko&t=143s) Uh, we wouldn't call this as a platform at this time.

[2:28](https://www.youtube.com/watch?v=n9rjuBuShko&t=148s) Very soon we realize, uh, there are some limitation for this simple approach, especially it lacks the capability to inject our rich data into the product experience. Then in the mid-2023, we started to develop the second generation of the GenAI product.

[2:47](https://www.youtube.com/watch?v=n9rjuBuShko&t=167s) Uh, internally, we call it Copilot or Coach. Here we're showing one popular such experience on LinkedIn right now. Uh, basically it looks at, uh, your profile and the job description and then, uh, use, uh, some RAG process to give you personalized recommendation on if you are a good fit to the job.

[3:10](https://www.youtube.com/watch?v=n9rjuBuShko&t=190s) At this time, we started to build, uh, some platform capability. Uh, specifically in the center of our platform, we build the, uh, Python SDK on top of the popular, uh, LangChain framework to orchestrate LLM calls.

[3:27](https://www.youtube.com/watch?v=n9rjuBuShko&t=207s) And it also provide the key value to integrate with our large-scale infrastructure, uh, in this SDK, so our developers can easily assemble an application. We started to unify the tech stack at this stage because we realized it's really costly to transfer the Python prompt into the Java world, not to mention the error during this process.

[3:51](https://www.youtube.com/watch?v=n9rjuBuShko&t=231s) We started to invest on the prompt management or prompt source of truth. This is a sub-module at this stage, uh, to help developers to version their prompt and to provide some structure around their meta prompt.

[4:07](https://www.youtube.com/watch?v=n9rjuBuShko&t=247s) Uh, the most important piece I'd like to call out here is conversational memory. Uh, this is, uh, infrastructure to help to keep track of the LLM interactions and retrieval content and then inject those content into the final product.

[4:24](https://www.youtube.com/watch?v=n9rjuBuShko&t=264s) It will help us to build this kind of conversational, uh, bot.

[4:31](https://www.youtube.com/watch?v=n9rjuBuShko&t=271s) Now, uh, zooming to this year, uh, actually in the last year, uh, we launched our first ever, uh, real multi-agent system called, uh, LinkedIn Hiring Assistant. Uh, this is, uh, multi-agent systems to help our recruiters to do their work, uh, efficiently, especially it automate several tedious task, uh, normally recruiter need to do manually, like,

[4:55](https://www.youtube.com/watch?v=n9rjuBuShko&t=295s) um, uh, post the job, um, and evaluate hundreds of candidates, then, uh, reach out to them.

[5:06](https://www.youtube.com/watch?v=n9rjuBuShko&t=306s) Our platform also start to evolve into the agent platform. Uh, from the framework side, we extend the support of the Python SDK into a more large-scaled, uh, distributed agent orchestration layer.

[5:21](https://www.youtube.com/watch?v=n9rjuBuShko&t=321s) It will handle the distributed agent execution and also handle the more complicated scenarios like retry logic and the traffic shift. Uh, for folks who build agent, uh, I think you probably know the skills or APIs are one key aspect of the agent because we expect, uh, this agent to perform some action.

[5:45](https://www.youtube.com/watch?v=n9rjuBuShko&t=345s) One investment we did at this, uh, time is around the skill registry. Basically, we have a set of tools, uh, to help our developers to publish their API into this centralized skill registry.

[5:59](https://www.youtube.com/watch?v=n9rjuBuShko&t=359s) This skill registry can handle the skill discovery problem-

[6:03](https://www.youtube.com/watch?v=n9rjuBuShko&t=363s) Skill invocation problem. So in your application, it's actually very easy to call the API to perform some task.

[6:11](https://www.youtube.com/watch?v=n9rjuBuShko&t=371s) Another key component, uh, we invest at this stage is on the memory. In addition to the conversational memory, we extend, uh, its capability into the experiential memory. Essentially, it's a memory storage to, uh, extract and analyze, and infer the contextual knowledge from the interaction between the agent and our user.

[6:33](https://www.youtube.com/watch?v=n9rjuBuShko&t=393s) We also organize this memory into different layers, including the, um, uh,

[6:40](https://www.youtube.com/watch?v=n9rjuBuShko&t=400s) working memory, long-term memory, collective memories. Uh, this can help our agent to be aware of the surrounding content. Uh, lastly, at this, uh, time, we also realize the observability is super important because agent, uh, one key aspect to define agent is autonomous, right?

[6:59](https://www.youtube.com/watch?v=n9rjuBuShko&t=419s) Uh, because agent can decide what API they can call, uh, what LL, uh, uh, LLM they need to call. So it's actually very hard to predict its behavior. So we started to invest on the observability.

[7:13](https://www.youtube.com/watch?v=n9rjuBuShko&t=433s) Uh, particularly, we build our in-house solution on top of the OTEL to keep track very low level granularity of the, uh, telemetry data. So we can use this data to replay the, uh, agent call.

[7:27](https://www.youtube.com/watch?v=n9rjuBuShko&t=447s) And we also add a actual layer of the analytics on top of it. So we can use that to guide the future optimization of our agent systems.

[7:37](https://www.youtube.com/watch?v=n9rjuBuShko&t=457s) Let's put together all the components we build for this platform. Uh, we can classify them into four layers, basically, including the orchestration, prompt engineering, tools and the skills invocation, content, and the memory, uh, management.

[7:54](https://www.youtube.com/watch?v=n9rjuBuShko&t=474s) Uh, of course, that's not everything in the LinkedIn GenAI ecosystem. Uh, in addition, we have our sister teams to build out the modeling layer, like fine-tune the open source models, responsible AI layers to make sure the agent is behave according to our policy and standard, and also the, uh, AI platform or machine learning infrastructure team to host

[8:19](https://www.youtube.com/watch?v=n9rjuBuShko&t=499s) those models. The key value proposition for this, uh, GenAI platform is actually to, uh, be the unified interface for this complex ecosystem.

[8:34](https://www.youtube.com/watch?v=n9rjuBuShko&t=514s) So our developers don't need to necessarily understand all those individual box when they build, uh, their application. Instead, they can leverage our platform to quickly access to this entire ecosystem.

[8:47](https://www.youtube.com/watch?v=n9rjuBuShko&t=527s) Uh, for example, uh, in our SDK, the developer can just switch one parameter in the one line of the code to switch from the OpenAI model to our on-prem model.

[9:00](https://www.youtube.com/watch?v=n9rjuBuShko&t=540s) Of course, they still need to do the prompt engineering, but that reduce a lot of the complexity on the infrastructure integration phase.

[9:09](https://www.youtube.com/watch?v=n9rjuBuShko&t=549s) Uh, last but the most important is because of this is a centralized platform, uh, it provide a place to enforce the best practice and governance. So we can make sure our developers are building the applications efficiently but also responsibly.

[9:28](https://www.youtube.com/watch?v=n9rjuBuShko&t=568s) As you can see from our journey, we actually started to build this, uh, platform piece by piece, and then this platform start to emerge. If we take one step back and think, oh, do we really need this platform at this time?

[9:41](https://www.youtube.com/watch?v=n9rjuBuShko&t=581s) Especially there are lots of, uh, uh, uh, vendor product, uh, on this space. Shall we buy it, build it? And why do we need to buy it or build it?

[9:51](https://www.youtube.com/watch?v=n9rjuBuShko&t=591s) Uh, here are some thoughts. Um, the short answer is yes. The reason behind it is, uh, we feel like, uh, GenAI is a totally different and new AI systems compared to the traditional AI systems.

[10:04](https://www.youtube.com/watch?v=n9rjuBuShko&t=604s) So in the traditional AI systems, there's a clear cutoff between the, uh, AI model optimization phase and the model serving phase. So AI engineers and the product engineers can operate in two different tech stack.

[10:19](https://www.youtube.com/watch?v=n9rjuBuShko&t=619s) Uh, they usually don't, uh, need to, uh, work on the same code base. But in the GenAI systems, what we're seeing is this line between the optimization phase and the serving phase disappear.

[10:33](https://www.youtube.com/watch?v=n9rjuBuShko&t=633s) Basically, everyone is AI engineer who can optimize the overall system performance. This actually create the new challenge of the tooling and the best practice in the company.

[10:47](https://www.youtube.com/watch?v=n9rjuBuShko&t=647s) Essentially, we think these GenAI systems or, uh, agent systems is a compound AI system. Here, we borrow the definition from Berkeley AI Research Lab. A compound AI system can be defined as a system which tackles AI tasks using multiple interacting components, including multiple calls to models, retrievers or external tools.

[11:10](https://www.youtube.com/watch?v=n9rjuBuShko&t=670s) As you can see, this is actually skill gap across AI engineer and product engineer. And I believe this, uh, GenAI platform is trying to bridge this gap.

[11:23](https://www.youtube.com/watch?v=n9rjuBuShko&t=683s) To summarize, uh, we believe this platform is critical for your success, mainly because it can bridge the skill gaps between those two group of engineers.

[11:34](https://www.youtube.com/watch?v=n9rjuBuShko&t=694s) Okay. Let's say if you want to build this, uh, platform in your company and how to hire it is a frequent question, uh, I heard. Um, I basically look into, uh, my great engineering team and, uh, extract all the key qualifiers from those top engineers.

[11:54](https://www.youtube.com/watch?v=n9rjuBuShko&t=714s) And, uh, I put all the qualifications here. Uh, the ideal candidate in this team is a strong software engineer, uh, who can build infrastructure integration. They have a good developer, uh, PM skills to design the interface.

[12:12](https://www.youtube.com/watch?v=n9rjuBuShko&t=732s) Uh, ideally, they have the AI and the data science background to understand the latest techniques. They are the people who can learn from the latest techniques, but at the same time, they are hands-on.

[12:26](https://www.youtube.com/watch?v=n9rjuBuShko&t=746s) Unfortunately, [chuckles] it's really hard to get those candidates. If you get them, uh, it's probably worth more than a unicorn. [laughs]

[12:33](https://www.youtube.com/watch?v=n9rjuBuShko&t=753s) Realistically, we are making multiple trade-offs in the hiring. Uh, here are some principle, uh, we follow, and it's actually working pretty well. Want to share here.

[12:45](https://www.youtube.com/watch?v=n9rjuBuShko&t=765s) In terms of the core skills, uh, we usually prioritize the stronger software engineer skills over the AI expertise. This might be controversial, but, uh, uh, uh, we can discuss if you are interested.

[13:01](https://www.youtube.com/watch?v=n9rjuBuShko&t=781s) Second is, instead of hiring for experience or degrees, we hire for the potential because this field is evolving so fast, most of the experience might be outdated.

[13:14](https://www.youtube.com/watch?v=n9rjuBuShko&t=794s) In case you won't be able to find a single engineer with all the qualifications we're showing here, uh, the way we are solving this problem is to hire or diversify the team.

[13:26](https://www.youtube.com/watch?v=n9rjuBuShko&t=806s) So, so for example, uh, in our team, we have some full stack software engineers. We have, uh, data scientists. We have AI engineers and, uh, data engineers. We also have, uh, fresh grads, uh, from the top research university and also, uh, some people from the startup background.

[13:49](https://www.youtube.com/watch?v=n9rjuBuShko&t=829s) And then we put them together, uh, into the project. What we're seeing is, based on those collaboration, those strong engineers start to pick up new skills in the project, and very soon, they started to grow into these ideal candidates.

[14:04](https://www.youtube.com/watch?v=n9rjuBuShko&t=844s) Uh, lastly is, uh, want to emphasize is, uh, the critical thinking. Uh, one constant topic, uh, in our team meeting is, uh, no matter what we're building right now, it will be outdated within a year or even less than six month.

[14:20](https://www.youtube.com/watch?v=n9rjuBuShko&t=860s) So we consistently evaluate the latest open source package, talking with vendors, and deprecate our solution more proactively.

[14:30](https://www.youtube.com/watch?v=n9rjuBuShko&t=870s) Cool. Let's talk about the sa-- uh, some, uh, key takeaways, uh, especially on the tech stack choice. If possible, we strongly recommend Python. We started with Java and Python.

[14:43](https://www.youtube.com/watch?v=n9rjuBuShko&t=883s) Uh, there are some back and forth of the debate internally, but finally, we picked Python, and I think that's the right choice, especially most research and open source, uh, are in this space.

[14:55](https://www.youtube.com/watch?v=n9rjuBuShko&t=895s) Based on our experience, it's also scalable. In terms of the, uh, key components you want to build in this platform, the first one is a prompt source of truth.

[15:06](https://www.youtube.com/watch?v=n9rjuBuShko&t=906s) Prompt in some way is like a traditional model parameters. You want to have a really robust systems to version control your prompt. This is really, really critical for the operational stability.

[15:18](https://www.youtube.com/watch?v=n9rjuBuShko&t=918s) You don't want accidentally edit your prompt in production and, uh, see some really side effect.

[15:25](https://www.youtube.com/watch?v=n9rjuBuShko&t=925s) Second key component is on the memory. I think in today's, uh, meeting, uh, I-I-I mean, today's talk, someone already talked about it. Memory is a really key component to inject your rich data into the agent experience.

[15:40](https://www.youtube.com/watch?v=n9rjuBuShko&t=940s) Lastly, in the agent era, uh, one key new component we are building is on the uplifting our APIs into skills which can be called from the agent easily. So you can, uh, build some surrounding tooling and infrastructure to support this need.

[16:02](https://www.youtube.com/watch?v=n9rjuBuShko&t=962s) All right. Let's talk about how to, uh, scale this solution and got-- get it adopted. Uh, from our experience, instead of trying to build this full-fledged platform at the beginning, try to solve immediate need.

[16:16](https://www.youtube.com/watch?v=n9rjuBuShko&t=976s) For example, we started with a simple Python library to support orchestration. Then we started to grow into all the components we're seeing here. Second is, uh, focus on the infrastructure and the scalable solution.

[16:31](https://www.youtube.com/watch?v=n9rjuBuShko&t=991s) At LinkedIn, we actually have a pretty good success story by leveraging our, uh, messaging infrastructure, uh, to be as a memory layer. Uh, it's both cost efficient and scalable.

[16:44](https://www.youtube.com/watch?v=n9rjuBuShko&t=1004s) Lastly is, uh, focus on the developer experience. By the end of the day, this platform is trying to help developer to be as productive as possible. Their adoption is a key for the success.

[16:57](https://www.youtube.com/watch?v=n9rjuBuShko&t=1017s) If you can design this platform, please focus on, uh, how to align your technology with their existing, uh, workflow, so it will ease adoption and, uh, be more successful.

[17:11](https://www.youtube.com/watch?v=n9rjuBuShko&t=1031s) Uh, we actually have lots of low-level details on the technical side. Uh, if you are interested, please check out our engineering blog post, uh, on LinkedIn, uh, by Karthik, Sandeep, and myself.

[17:24](https://www.youtube.com/watch?v=n9rjuBuShko&t=1044s) Uh, with that, uh, thank you for your attention, and, uh, if you are having more questions, happy to answer that after the talk. Thank you. [upbeat music]
