# No-code Fine-tuning: Mark Hennings

Mark Hennings

AI Engineer Summit 2023

[Talk page](https://ai.engineer/talks/LksXn4CLC0g-no-code-fine-tuning) · [Video](https://www.youtube.com/watch?v=LksXn4CLC0g)

## No-code Fine-tuning

Mark Hennings shows how a team can turn examples into a specialized language model, evaluate it against a prompt, and build a press release workflow around facts.

Familiarity with prompting a language model is helpful; no programming or model-training experience is required.

### What fine-tuning specializes

[0:22](https://www.youtube.com/watch?v=LksXn4CLC0g&t=22s)

How do you write a product description, normalize messy text, or decide which support issue deserves attention first? These are recurring business tasks, but specifying the right behavior as rules can be difficult. **Fine-tuning** means training a foundation model for a specialized task. Mark Hennings opens with three overlapping groups:

- **Writing:** emails, blog articles and product descriptions.
- **Transforming data:** scrubbing fake emails from a list, extracting or normalizing information, translating, paraphrasing and rewriting.
- **Making judgments:** qualifying sales leads, ranking support priorities, detecting fraud and flagging inappropriate content.

The common challenge is interpreting language with enough nuance to produce a useful result. Hennings argues that language models handle this more easily than traditional programming or rule-based approaches.

![Slide lists eight specialized tasks, with a callout describing them as hard for rule-based systems or programming and easy for large language models.](https://ai.engineer/api/talk-essay-media/cbacabf2eaee7b8ccd85ecc544a1461991dcf5e970a7afa9ea61827acfc28f6c)

Specialized tasks for language models, from writing copy to detecting fraud.

A prompt can already accomplish many of these tasks. The question is what changes when recurring instructions and examples become training data for a specialized model.

### Speed, cost and shorter prompts

[1:23](https://www.youtube.com/watch?v=LksXn4CLC0g&t=83s)

The proposed efficiency gain starts with **matching the prompted model’s quality using a lighter model**. If training teaches that smaller model the required behavior, inference can become faster and cheaper. Removing repeated instructions also shortens inputs, leaving more context space for completions. Training examples provide another place to express edge cases and let teammates contribute improvements.

Hennings also asserts that fine-tuning is naturally resistant to prompt injection, but he does not demonstrate an attack or a defense. Ordinary task fine-tuning should not be treated as a security guarantee.

For speed, Hennings compares **GPT-4** with **GPT-3.5** on the OpenAI API, then gives a cost example comparing GPT-4 with fine-tuned GPT-3.5. These are figures from the historical 2023 demonstration, not current model or pricing guidance.

| Comparison | Reported result |
| --- | --- |
| OpenAI API response time per token | Hennings reports approximately 196 ms for GPT-4 and 73 ms for GPT-3.5. |
| GPT-4 versus fine-tuned GPT-3.5 cost | Hennings reports 88.6% savings in his example. |

He describes the latency difference as roughly three times faster. The talk does not specify the model snapshots, workload or measurement procedure, and it does not supply the token counts or training-cost allocation needed to reproduce the savings calculation.

The prompt-compression example is a blog writer. Its engineered prompt specifies the topic, writing approach, tone, desired behavior and things to avoid. With fine-tuning, examples teach the recurring writing behavior; the request can concentrate on the topic that changes from one article to the next.

Hennings describes this particular blog-writing prompt as 90% shorter. The reduction depends on how much of the original prompt consists of reusable instructions. The useful separation is between behavior shared across requests and information unique to the current request.

### Training data as a shared working surface

[2:57](https://www.youtube.com/watch?v=LksXn4CLC0g&t=177s)

Imagine a GitHub repository whose entire codebase lives in one file. Hennings uses that as an analogy for a team maintaining one enormous prompt: everyone must work on the same artifact. Fine-tuning offers something closer to a repository with multiple files, where contributors can work on different pieces. Those pieces are training examples. Team members can add, edit and improve examples in a shared dataset, which then feeds the fine-tuned model.

![Diagram contrasts four people pointing toward One Epic Prompt with four contributors above training data that feeds into a Fine-tuned Model.](https://ai.engineer/api/talk-essay-media/be97b2dbcf1ec2f7f0ea4b7024bed3456b0040e78cd519fea29daadbd0731e7e)

Team contributions to one epic prompt versus training data feeding a fine-tuned model.

This makes the dataset a collaborative layer for specifying behavior. The reason to adopt it remains conditional: the resulting model needs to produce equal or better output. Hennings then turns to the practical obstacle—fine-tuning still looks like a developer’s job.

### Lowering the technical and data barriers

[3:42](https://www.youtube.com/watch?v=LksXn4CLC0g&t=222s)

A typical fine-tuning tutorial can involve provisioning GPU servers for training and inference, formatting data with ad hoc Python scripts, configuring parameters and making API calls. Hennings’s proposal is to automate that operational work behind a user interface. The person who knows what a good answer looks like should be able to contribute without owning every infrastructure step.

Hennings says 20 examples can be enough to start a fine-tuning experiment. They can even be handwritten, avoiding the initial burden of assembling the thousands of examples he associates with traditional machine learning. Starting an experiment is distinct from establishing production quality; that requires evaluation.

His intuition comes from **few-shot learning**. A prompt might contain five examples of the desired behavior. A training dataset can instead contain 20 or 100, without placing all those examples into each request. This extends the amount of example-based guidance available to the model. Hennings presents improved task fit from additional examples as an intuition, rather than a measured result.

### Prototype, evaluate and learn from production

[4:56](https://www.youtube.com/watch?v=LksXn4CLC0g&t=296s)

Prompt engineering remains the starting point in the proposed development lifecycle:

1. **Prototype with a prompt.** Validate the concept and use the prompted model to help create an initial training dataset.
2. **Fine-tune and evaluate.** Check whether the trained model improves on the prompted version, then test which models can achieve the required performance.
3. **Deploy after evaluation.** Put the selected fine-tuned model into production.
4. **Feed experience back into training.** Capture user feedback and log real examples, then incorporate useful cases into the dataset for subsequent improvement.

The loop connects the initial specification to the situations users actually encounter.

Developers can operate this process, but they need not be its only participants. People who understand the task can help with prompting, example creation and fine-tuning. Hennings introduces [Entry Point](https://www.entrypointai.com/), identifying himself as a co-founder, as tooling intended to make that participation practical.

### Building the Press Release Writer dataset

[6:02](https://www.youtube.com/watch?v=LksXn4CLC0g&t=362s)

The demonstration opens the `Press Release Writer` project in Entry Point, containing 20 examples. Hennings began with the desired outputs: he found 20 strong press releases through articles collecting examples of effective releases. That gave him writing to learn from, but not complete input–output pairs.

To supply the missing inputs, he used GPT-4 in ChatGPT to work backward from each release. The task was to extract the list of facts a professional writer would need to produce it. Each resulting pair therefore associates a factual brief with a finished press release.

**Facts are the input; polished prose is the output.** Hennings chooses that boundary because he does not want to rely on the model for factual knowledge. The intended result is a strong first draft built from supplied information. The examples teach how to turn that information into a release.

### From structured fields to a training job

[6:52](https://www.youtube.com/watch?v=LksXn4CLC0g&t=412s)

Entry Point exposes the training data as structured records. When a CSV is imported, each column becomes a field; this project has fields for the facts and the press release. Keeping them visible makes it easier to inspect what will enter the model’s training data.

Fields are inserted into input and output templates using [Handlebars](https://handlebarsjs.com/guide/), much like inserting a contact’s first name into a personalized email. With illustrative field identifiers `facts` and `press_release`, the input template can be:

```handlebars
{{facts}}
```

The corresponding output template can be:

```handlebars
{{press_release}}
```

Applying those templates to each record supplies its factual brief as the input and its existing release as the target output. The templates define the format once; the records provide the changing content.

Fine-tuning does not require abandoning instructions. The demonstrated **GPT-3.5 Turbo** workflow also supports a system prompt. Hennings uses that to describe a hybrid: examples teach the task while a system instruction provides additional guidance, including when the dataset is small.

Once the dataset and templates are ready, the interface handles the training setup:

1. Open `Fine-tunes` and select `Add`.
2. Choose the model and platform; Entry Point supports multiple platforms.
3. Review the token count and estimated cost.
4. Press `Start` to launch the job.

Hennings quotes an estimated training cost of about $1 for the demonstrated job. The talk does not provide the token counts, selected model or training settings needed to reproduce that estimate. He then switches to an already-trained model in Entry Point Playground, rather than waiting for the new job to finish.

### Iterating on facts instead of prompt boilerplate

[8:21](https://www.youtube.com/watch?v=LksXn4CLC0g&t=501s)

In the Playground, Hennings supplies facts he wrote about the [AI Engineer Summit](https://ai.engineer/summit/2023). The trained model generates a draft with a title and recognizable press-release formatting. The demonstration shows the facts-to-draft workflow; it does not establish held-out quality or factual accuracy.

The generated draft becomes material for another pass:

1. Write the list of facts.
2. Generate the article.
3. Read the draft and use it to identify ideas or refinements.
4. Revise the facts and generate again.

This is the authoring loop Hennings finds useful. Training data carries recurring writing conventions, so each iteration can focus on what the release should say. The user works on the factual brief while the trained behavior supplies much of the writing boilerplate.

Hennings closes by mentioning Entry Point’s data-synthesis features and tools for comparing fine-tuned models, neither of which is demonstrated here, and directs viewers to the product. The completed walkthrough ends with a more focused writing process: inspect the draft, improve the facts, and let the examples carry the reusable style.

## Resources

- [Entry Point AI](https://www.entrypointai.com/) — The platform demonstrated in the talk, now offering fine-tuning, synthetic-data generation and data transformation workflows.
- [Hennings's guide to no-code fine-tuning](https://www.entrypointai.com/blog/fine-tune-large-language-models/) — A 2023 walkthrough of CSV imports, field templates, training, validation and the Entry Point playground.
- [OpenAI's GPT-3.5 Turbo fine-tuning announcement](https://openai.com/index/gpt-3-5-turbo-fine-tuning-and-api-updates/) — The original announcement documents narrow-task results, prompt compression and launch-era training and inference prices.
- [Hennings on the GPT-3.5 Turbo fine-tuning format](https://www.entrypointai.com/blog/gpt-3-5-turbo-fine-tuning-update-for-developers/) — A contemporary explanation of chat-format training examples and combining system instructions with fine-tuning.
- [Handlebars template syntax](https://handlebarsjs.com/guide/) — Official guide to substituting field values into templates, with interactive examples.
- [Training models to prioritize trusted instructions](https://openai.com/index/the-instruction-hierarchy/) — Later research explains how targeted instruction-hierarchy training improves resistance to prompt injection.

## Slides

### Diamond Sponsors

Fixie and Supabase are listed as diamond sponsors.

[0:09](https://www.youtube.com/watch?v=LksXn4CLC0g&t=9s) [Diamond sponsors slide with Fixie and Supabase logos.](https://assets.ai.engineer/sites/aiecode2025/sha256/b3edd6c6072875ff1282697e19201c1cfe0bada6599bffeb282f6fe6e5368755.webp)

DIAMOND SPONSORS

FIXIE
supabase

### No-code Fine-tuning

Mark Hennings introduces fine-tuning large language models without code.

[0:15](https://www.youtube.com/watch?v=LksXn4CLC0g&t=15s) [Entry Point AI title slide for No-code Fine-tuning.](https://assets.ai.engineer/sites/aiecode2025/sha256/7b1c871a19a0a03fa51dec61eedfb5261acc9eb48f9b7f3f8fb4a8a56d0620d1.webp)

No-code Fine-tuning

Entry Point AI
entrypointai.com

### What Is Fine-tuning?

Mark Hennings defines fine-tuning as training a foundation model for a specialized task, such as extracting data, writing copy, or flagging inappropriate content.

[0:34](https://www.youtube.com/watch?v=LksXn4CLC0g&t=34s) [Fine-tuning is defined as training a foundation model for a specialized task.](https://assets.ai.engineer/sites/aiecode2025/sha256/a4c270d2945ced725506f0498368e632c25d35661253c73d1ec6e4d80a8b88bd.webp)

For our purposes, fine-tuning is:

Training a foundation model for a specialized task.

### Why Fine-tune?

Hennings argues that a lighter fine-tuned model can match prompt-based quality faster and more cheaply. He also cites shorter prompts, edge-case coverage, team collaboration, and resistance to prompt injection.

[1:36](https://www.youtube.com/watch?v=LksXn4CLC0g&t=96s) [The slide lists speed and cost, shorter prompts, training examples and resistance to prompt injection as benefits.](https://assets.ai.engineer/sites/aiecode2025/sha256/136de4dc21cd9956caf42f8f88b93cab9057ca592c448b181c26aea91411d7cb.webp)

Why fine-tune?

Because it’s awesome.

Faster and cheaper because you can train a lighter model to match the quality.

Reduces the size of your prompts, allowing for longer completions.

Training examples allow you to cover edge cases and collaborate better as a team.

Naturally resistant to prompt injection attacks.

### How Much Faster?

The chart compares OpenAI API response times: GPT-3.5 takes 73 ms per token versus GPT-4’s 196 ms, about 2.7 times faster.

[1:49](https://www.youtube.com/watch?v=LksXn4CLC0g&t=109s) [A bar chart compares response times per token: 196 milliseconds for GPT-4 and 73 milliseconds for GPT-3.5.](https://assets.ai.engineer/sites/aiecode2025/sha256/89304b868007649c0f7f27b9dc63749c3e12242abbf392f5db1a427bebd43fbd.webp)

How much faster? 🐇

OpenAI API Response Times (Milliseconds per Token)

GPT-4: 196ms
GPT-3.5: 73ms

Source: https://www.taivo.ai/__gpt-3-5-and-gpt-4-response-times/

### How much cheaper?

In this 1M-request example, fine-tuned GPT-3.5 costs $3,600 versus $31,500 for GPT-4: 88.6% savings, with a lower price and shorter prompt.

[2:06](https://www.youtube.com/watch?v=LksXn4CLC0g&t=126s) [A chart compares $31,500 for GPT-4 with $3,600 for fine-tuned GPT-3.5 over one million requests, showing 88.6% savings.](https://assets.ai.engineer/sites/aiecode2025/sha256/216bbb3c74c196b9fba323d77c7ef20b088e56915be1576606a14f79100572c1.webp)

How much cheaper? 💰

1M Requests

GPT-4: $31,500
GPT-3.5 Fine-tuned: $3,600

⭐ 88.6% savings

Lower price & shorter prompt for fine-tuned model

### How much shorter prompts?

The fine-tuned model learns the writing style from examples, so the prompt can contain just the topic. Hennings estimates this example is 90% shorter.

[2:22](https://www.youtube.com/watch?v=LksXn4CLC0g&t=142s) [A detailed engineered blog-writing prompt is compared with the short topic prompt for a fine-tuned model.](https://assets.ai.engineer/sites/aiecode2025/sha256/fc29ff04f6a7584557a8fd9b1290b9a20e43c22c2e4bf45ed92c80e518cf6bcc.webp)

How much shorter prompts?

Engineered Prompt

Write a blog post on “How to Fine-tune a LLM.”

Use a professional tone with 10th grade language. Include some analogies to aid understanding but not too many. Write in short paragraphs and include headings to separate content into logical sections and paragraphs.

Conclude the blog post with a compelling call-to-action for the Entry Point AI platform, a fine-tuning software platform for large language models.

Prompt for Fine-tuned Model

How to Fine-tune a LLM

### How much better for teams?

Instead of everyone working on one epic prompt, team members can add and edit training examples that feed a fine-tuned model.

[3:10](https://www.youtube.com/watch?v=LksXn4CLC0g&t=190s) [A shared epic prompt is compared with team contributions to training data that feeds a fine-tuned model.](https://assets.ai.engineer/sites/aiecode2025/sha256/811dc168aa367ba0c1b328a7bb741edc3f63fc1197777715869bb9c69511f9ff.webp)

How much better for teams?

One Epic Prompt

Training Data

Fine-tuned Model

### Why Devs Handle Fine-tuning Today

Fine-tuning looks like a developer job because it involves GPU servers, data formatting, parameters and API calls. Hennings asks whether a user interface could automate those steps.

[3:54](https://www.youtube.com/watch?v=LksXn4CLC0g&t=234s) [Three bullets describe setting up GPU servers, formatting training data with Python scripts, and configuring parameters and API calls.](https://assets.ai.engineer/sites/aiecode2025/sha256/55904fcf280315e5bcd3b24a8bacd48166be4515208a0ee144c9ae6db8cf5e32.webp)

Why Devs Handle Fine-tuning Today

• Spinning up GPU servers for training and inference
• Formatting training data with ad hoc Python scripts
• Configuring parameters and making API calls

[4:00](https://www.youtube.com/watch?v=LksXn4CLC0g&t=240s) [Slide lists the technical tasks involved in fine-tuning and asks whether a user interface could automate them.](https://assets.ai.engineer/sites/aiecode2025/sha256/4a7f3c1adfc6d3a818c318b5ed3b665ae4ed3b59585a2534a5fc459e4176c71a.webp)

Why Devs Handle Fine-tuning Today

• Spinning up GPU servers for training and inference
• Formatting training data with ad hoc Python scripts
• Configuring parameters and making API calls

🤔 Why can’t all that be automated with a convenient user interface?

### Proposed LLM Dev Lifecycle

The proposed workflow moves from prompt engineering through fine-tuning to production, then feeds logged examples back into the dataset. Non-developers can participate in the first two stages.

[5:45](https://www.youtube.com/watch?v=LksXn4CLC0g&t=345s) [Diagram moves from prompt engineering to fine-tuning and production, with feedback from production to fine-tuning.](https://assets.ai.engineer/sites/aiecode2025/sha256/eb66612deae5b192f9c811d99f16c2b7370323c44e51957e9c3377a0feae40c9.webp)

Proposed LLM Dev Lifecycle

Everybody else
Developers

Prompt Engineering → Fine-tuning ↔ Production

Create a prototype and validate the concept
Evaluate models and build the dataset
Capture feedback and log examples

### Entry Point Tooling

Entry Point’s interface brings field definitions, labeled examples, and fine-tune setup together. The example classifies support messages by priority, with OpenAI and davinci selected for the model setup.

[5:56](https://www.youtube.com/watch?v=LksXn4CLC0g&t=356s) [Slide shows interface examples for defining fields, editing training examples and starting a fine-tune.](https://assets.ai.engineer/sites/aiecode2025/sha256/71c14c990769a991d6b5ddbb250210135905e81e1fdff99662d636f2507d0ec5.webp)

Entry Point offers modern tooling to make it easy.

Fields
+
Name | Target | Reference | Type | Actions
Priority | Completion | priority | Predefined Options | Edit
Message | Prompt | message | Text | Edit

Edit Example
Training | Validation
Prompt
Message
The website is down and I have orders I need to fulfill ASAP. Please help!
Completion
Priority
urgent

I am trying to learn more about the built-in CRM integrations. Do you have any resources or videos I could watch?
low
4 min ago

The website is down and I have orders I need to fulfill ASAP. Please help!
urgent
5 min ago
23 tokens

Platform
OpenAI
Base Model
davinci
Start

### Press Release Writer Overview

Hennings built 20 examples from press releases he found online, using ChatGPT-4 to generate the missing input facts. The overview shows one example alongside recent fine-tune activity.

[6:06](https://www.youtube.com/watch?v=LksXn4CLC0g&t=366s) [Project overview displays a sample press release prompt and recent project activity.](https://assets.ai.engineer/sites/aiecode2025/sha256/1ea6b40586edf0be4327be734481ded0ebab2f05f9a10a7ff9ad10ad867ced14.webp)

Overview
Generator
20 examples
2 fields

Example Roulette

You are a professional writer who will expertly craft and polish an official press release from a list of facts. Write it in third-person perspective.

Facts:
Titleist introduced the New AVX Golf Ball on January 22, 2020, in FAIRHAVEN, Mass.
This golf ball is designed to boost ball speed, increase distance, and enhance short-game spin.
It was first launched in the spring of 2018.
The AVX is positioned as an alternative to Titleist’s Pro V1 and Pro V1x models.
It boasts a larger, redesigned low-compression core.
The high flex casing layer has been revamped to improve ball speed and reduce long game spin.
Titleist Golf Ball R&D formulated a thinner cast urethane cover specifically for the AVX.
The increased core size provides a softer feel for golfers.
The AVX is available in both white and high optic yellow.
It can be found in Australian and New Zealand golf shops.
Frederick Waddell, Senior Product Manager at Titleist Golf Balls, commented on these improvements and golfer feedback.
[illegible]
7 days ago
1,475 tokens

You are a professional writer who will expertly craft and polish an official press release from a list of facts. Write it in third-person perspective.

Facts:

Recent Activity

10 min ago
Mark Hennings started the Mint Green Mink fine-tune with 20 examples.

4 days ago
Mark Hennings started the Black Porpoise fine-tune with 20 examples.

7 days ago
Mark Hennings started the Press Release Generator fine-tune with 20 examples.

7 days ago
Mark Hennings started the Mint Green Caribou fine-tune with 20 examples.

7 days ago
Mark Hennings created 3 fields.

7 days ago
Mark Hennings created this project.

Interface text:
Projects / Press Release Writer / Overview

### Press release examples

The Examples page exposes the data used for fine-tuning: cards pair a press-release writing instruction with lists of facts and, in one visible example, the resulting release. Hennings says imported CSV columns become fields.

[6:51](https://www.youtube.com/watch?v=LksXn4CLC0g&t=411s) [Examples page showing press release training examples and filtering controls.](https://assets.ai.engineer/sites/aiecode2025/sha256/f961108248c8fb1ea76ab65c41b032dd202d66aa678f605bb985e2ccc6438dd6.webp)

Examples

ALL | TRAINING | VALIDATION
Search
Search Field
Sort By: Updated
Order: Descending
20 results

You are a professional writer who will expertly craft and polish an official press release from a list of facts. Write it in third-person perspective.

Facts:
Tracepoint is a company that deals with digital forensics and incident response (DFIR).
Booz Allen Hamilton acquired Tracepoint as part of its capital deployment plan.
In January 2021, Booz Allen initially invested in Tracepoint before the acquisition.
The exact terms of the acquisition deal were not publicly disclosed.
Tracepoint was co-founded in December 2019 in Fredericksburg, Virginia, by Plexos Group, Chris Salsberry, Brett Anderson, and Rob Driscoll.
Tracepoint specializes in assisting cyber insurance companies, lawyers, brokers, and clients in handling cyber incidents.
Booz Allen intends to merge Tracepoint with its Commercial Cyber division in early 2022.
This integration aims to establish a larger business focusing on incident response, [illegible]
7 days ago | 2,228 tokens

You are a professional writer who will expertly craft and polish an official press release from a list of facts. Write it in third-person perspective.

Facts:
Event Name: MIT Sloan CFO Summit
Date: November 18th, 2021
Location: Boston Marriott Newton
Theme: "Optimism and Excellence"
Chair: Jeremy Seidman
Event Founded in: 2002
Brings Together: Financial executives and MIT faculty
Virtual Participation Option Available
Speaker Line-up: CFOs from various companies including Blue Cross Blue Shield of Massachusetts, Bristol Myers Squibb, Desktop Metal, Hologic, Spotify, Stripe, Soundcloud, SQZ Biotech, ViacomCBS, Vimeo, Workhuman
Morning Fireside Chat: Dhivya Suryadevara (CFO, Stripe) in conversation with Bill [illegible]
7 days ago | 1,250 tokens

You are a professional writer who will expertly craft and polish an official press release from a list of facts. Write it in third-person perspective.

Facts:
The Certificate in Sustainable Marketing is a 3-module masterclass series.
Green jobs and skillsets are on the rise, according to the World Economic Forum’s Future of Job Report 2023.
The Direct Marketing Association of Northern California was founded to educate and connect marketing professionals in the San Francisco Bay Area and beyond.
The program, led by Gemma Butler, co-founder of Can Marketing Save the Planet Ltd., aims to make marketers part of the solution in creating sustainable markets and reshaping societal behaviors.
The masterclass series focuses on greenwashing and what marketers need to know.
The next certificate series begins on September 29, 2023.
Gemma Butler, who also hosts the podcast ‘Can Marketing Save the Planet,’ co-authored the book ‘Sustainable Marketing, How to Drive Profits with Purpose.’
[illegible]
7 days ago | 1,436 tokens

and alignment with off-peak, low-cost hours.
BMW, Ford, and Honda have direct communication with their EV customers.
The company will reduce EV customers’ personal carbon footprints while supporting grid resiliency.
https://csr.honda.com/environment/na-environmental-report/

Press release:
BMW, Ford and Honda Agree to Create ChargeScape, a New Company Focused on Optimizing Electric Vehicle Grid Services

ChargeScape will aim to unlock the full potential of electric vehicle (EV) technology through conveniently managed energy services never before possible with gasoline-powered vehicles — benefiting EV customers, electric utilities and automakers. BMW Group, Ford and Honda will leverage years of cross-industry research around the Open Vehicle-Grid Integration Platform (OVGIP) to create a single platform that will seamlessly connect electric utilities, automakers and their interested EV [illegible]
7 days ago | 2,462 tokens

You are a professional writer who will expertly craft and polish an official press [illegible]
You are a professional writer who will expertly craft and polish an official press [illegible]

Interface text:
Press Release ...
Dashboard
Integrations
Upgrade
Admin
Help
Mark’s Entry Point
Project
Overview
Examples 20
Fine-tunes 4
Playground
Synthesis
Settings
General
Template Chat
Fields 2
Import
Export
Projects / Press Release Writer / Examples

### Press release template

The template pairs a system instruction for third-person press releases with Handlebars fields for the facts input and press release output.

[7:13](https://www.youtube.com/watch?v=LksXn4CLC0g&t=433s) [Chat template with system instructions, a facts input, and a press release output field.](https://assets.ai.engineer/sites/aiecode2025/sha256/8a869b16c744ddf30ecac203ff4271cca6b3c78583f7efdc5f98cbb115415cfe.webp)

Template

Prompt
System
You are a professional writer who will expertly craft and polish an official press release from a list of facts. Write it in third-person perspective.

User input
Facts:
{{facts}}

Completion
Assistant output
Press release:
{{press_release}}

Interface text:
Projects / Press Release Writer / Template
STANDARD
CHAT
Prompt display: Plain Text
Completion display: Plain Text
Save

### Start a fine-tune

The demo selects OpenAI and GPT-3.5 Turbo for a fine-tune. The setup counts 31,585 training tokens across 20 examples and estimates a cost of about $1 and a time of 7 minutes.

[7:54](https://www.youtube.com/watch?v=LksXn4CLC0g&t=474s) [Fine-tune setup with platform and base model selectors, training counts, and cost and time estimates.](https://assets.ai.engineer/sites/aiecode2025/sha256/9226058a633e72ff5e68f81127f8c0005d96c5393f1c9c400fc1c6d298c4d7c5.webp)

Start a Fine-tune

Platform: OpenAI
Base Model: GPT-3.5 Turbo
Show Advanced
Start

Data Set | # of Examples | Total Tokens
Training | 20 | 31,585
Validation | 0 | 0

Estimates
Cost: $1.0[illegible]
Time: 7 minutes

Actual costs and times may vary. You are responsible for all charges. Validation examples are evaluated after fine-tuning and incur additional charges.

### Fine-tune jobs

A new GPT-3.5 Turbo job is preparing while four earlier jobs are complete. The presenter opens a completed model to try generating a press release in the playground.

[8:09](https://www.youtube.com/watch?v=LksXn4CLC0g&t=489s) [Fine-tune list showing one preparing job and several completed jobs.](https://assets.ai.engineer/sites/aiecode2025/sha256/48ac307194504ef2ff58cdb00ef1af818eec45ae61e0b65d72c09997e541923a.webp)

Fine-tunes
Train AI on your platform of choice.

Name | Score | Platform | # Examples | Model | Status
Salmon Weasel | | OpenAI | | gpt-3.5-turbo | Preparing, Just now
Mint Green Mink | 0% | OpenAI | 20 | ft:gpt-3.5-turbo-0613:simple-booth:press-release-writ:[illegible] | Completed, 7 min ago
ID: ftjob-[illegible]
Black Porpoise | 0% | OpenAI | 20 | ft:gpt-3.5-turbo-0613:simple-booth:press-release-writ:[illegible] | Completed, 4 days ago
ID: ftjob-[illegible]
Press Release Generator | 0% | OpenAI | 20 | ft:gpt-3.5-turbo-0613:simple-booth:press-release-writ:[illegible] | Completed, 7 days ago
ID: ftjob-[illegible]
Mint Green Caribou | 0% | OpenAI | 20 | ft:gpt-3.5-turbo-0613:simple-booth:press-release-writ:[illegible] | Completed, 7 days ago
ID: ftjob-[illegible]

Interface text:
Projects / Press Release Writer / Fine-tunes
+

### Fine-tune details

Black Porpoise was trained on 20 examples. Its template asks for a third-person press release from a list of facts; the demo moves toward the Playground to try it.

[8:13](https://www.youtube.com/watch?v=LksXn4CLC0g&t=493s) [Completed fine-tune details alongside its system, user, and assistant template.](https://assets.ai.engineer/sites/aiecode2025/sha256/847e9d37854973597093d683279ce353ee63914611d393db16c5efa9b2a56dfb.webp)

OPENAI GPT-3.5-TURBO
Black Porpoise

Details
Created: September 15, 2023
Training time: 5 minutes
# of examples: 20
Fine-tuned model: ft:gpt-3.5-turbo-0613:simple-booth:press-release-writ:[illegible]
OpenAI ID: ftjob-[illegible]
Training files: file-[illegible]
N epochs: 4
Prompt loss weight: 0.01

Template
System
You are a professional writer who will expertly craft and polish an official press release from a list of facts. Write it in third-person perspective.

User
Facts:
{{facts}}

Assistant
Press release:
{{press_release}}

Sharing Access
Create a link to share this model with others.
Enable

Interface text:
Projects / Press Release Writer / Fine-tunes / Black Porpoise
Playground

### Press release playground

The playground is ready to test the trained Black Porpoise model: enter facts on the left, then generate a press release in the completions area.

[8:17](https://www.youtube.com/watch?v=LksXn4CLC0g&t=497s) [Empty facts prompt alongside fine-tuned model settings.](https://assets.ai.engineer/sites/aiecode2025/sha256/0b2277f5cb5d141170e65783418d37cb6401a90251c2c044c21f26e60324943c.webp)

Playground

Prompt
List of Facts
Press Enter after each fact to start the next on a new line.

Completions

Settings
Advanced
Fine-tuned Model
Black Porpoise
ft:gpt-3.5-turbo-0613:simple-booth:press-release-writ:[illegible]
Temperature: 0.6
Max Tokens: 4096
Quantity: 1
Get completion

Interface text:
Projects / Press Release Writer / Playground
Reset

### Entering summit facts

Hennings enters summit facts and a speaker quote into the playground to generate a press release with a fine-tuned model.

[8:25](https://www.youtube.com/watch?v=LksXn4CLC0g&t=505s) [A list of AI Engineer Summit facts entered into the playground prompt.](https://assets.ai.engineer/sites/aiecode2025/sha256/c1dd1fa62ce7b3d08318d96bfe14e925ed3647b1381ff3c5500ba13d1d83017b.webp)

Playground

Prompt
List of Facts
The inaugural AI Engineer Summit was a huge success.
Hosted in San Francisco, Oct 8-10, 2023 at Hotel Nikko near historic Powell St Cable Car turnaround and BART station
Featured talks from leading AI companies & technologies, including Github, OpenAI, Amazon, Langchain, LlamaIndex.
Covering topics of highest relevance to AI Engineers including AI Agents, AI UX, AI Infra, AI Devtools and Codegen, LLMops, RAG/Vector DBs, Open Source AI, Multimodal APIs, & more
Over 500 developers and business executives attended.
Thousands of remote attendees.
Sessions streamed live.
Sponsored by companies that are pushing the boundaries of the AI Engineering ecosystem including: AutoGPT, Fixie, Supabase, and many others.
"For the first time in decades, a new kind of software engineer is emerging and this is the conference to explore it."
-Amjad Masad, Founder & CEO, Replit

Press Enter after each fact to start the next on a new line.

Settings
Advanced
Fine-tuned Model

Completions

Interface text:
Projects / Press Release Writer / Playground
Reset

### Generated press release

The fine-tuned model turns summit facts into a press release with a headline. Hennings describes reading the draft and revising the facts as an iterative workflow.

[8:41](https://www.youtube.com/watch?v=LksXn4CLC0g&t=521s) [Summit facts beside the visible portion of a generated press release.](https://assets.ai.engineer/sites/aiecode2025/sha256/583e64b9659559ded0befb5f2fa4e0bceb50dd1060975a651a3b5e3e633d7282.webp)

Prompt

List of Facts
The inaugural AI Engineer Summit was a huge success.
Hosted in San Francisco, Oct 8-10, 2023 at Hotel Nikko near historic Powell St Cable Car turnaround and BART station
Featured talks from leading AI companies & technologies, including Github, OpenAI, Amazon, Langchain, LlamaIndex.
Covering topics of highest relevance to AI Engineers including AI Agents, AI UX, AI Infra, AI Devtools and Codegen, LLMops, RAG/Vector DBs, Open Source AI, Multimodal APIs, & more
Over 500 developers and business executives attended.
Thousands of remote attendees.
Sessions streamed live.
Sponsored by companies that are pushing the boundaries of the AI Engineering ecosystem including: AutoGPT, Fixie, Supabase, and many others.
"For the first time in decades, a new kind of software engineer is emerging and this is the conference to explore it."
-Amjad Masad, Founder & CEO, Replit

Completions

Black Porpoise
Press Release:
AI Engineer Summit Draws Over 500 Attendees; First Event of Its Kind Showcases New and Emerging AI Engineer Role

SAN FRANCISCO, Oct. 17, 2023 /PRNewswire/ -- The inaugural AI Engineer Summit was a huge success, drawing over 500 developers and business executives to Hotel Nikko in San Francisco, California from October 8 to 10, 2023. Thousands of additional remote attendees joined via the event's live stream.

The AI Engineer Summit is a conference focused on the emerging profession of AI Engineering. The event was the first of its kind, featuring talks from some of the world's leading AI companies and technologies, including Github, OpenAI, Amazon, Langchain, LlamaIndex, and many others. The conference was organized around the themes that matter most to AI Engineers, including AI Agents, AI UX, AI Infra, AI Devtools and Codegen, LLMops, RAG/Vector DBs, Open Source AI, Multimodal APIs, and more.

"For the first time in decades, a new kind of software engineer is emerging and this is the conference to explore it," said Amjad Masad, Founder & CEO of Replit, and one of the Summit's opening keynotes.

The conference was organized by the team behind the popular AI Engineering publication, AI Engineer. AI Engineer is a community and newsletter for AI Engineers, with over 25,000 members.

"The AI Engineer Summit was a huge success," said AI Engineer Founder & CEO Logan Hanks. "We were thrilled to have so many talented AI Engineers join us in San Francisco, and to broadcast our event to thousands more on the live stream. We're already looking forward to the next one."

Interface text:
Playground
Reset
Press Enter after each fact to start the next on a new line.
Settings
Advanced
Fine-tuned Model
Black Porpoise
ft:gpt-3.5-turbo-0613:simple-booth:press-release-writ:[illegible]

## Full transcript

[0:00](https://www.youtube.com/watch?v=LksXn4CLC0g&t=0s) [upbeat music] Hey, my name is Mark Hennings.

[0:22](https://www.youtube.com/watch?v=LksXn4CLC0g&t=22s) I'm a serial entrepreneur, and I'm super excited to talk to you about fine-tuning large language models today without any code. So let's begin. For our purposes today, fine-tuning is training a foundation model for a specialized task.

[0:38](https://www.youtube.com/watch?v=LksXn4CLC0g&t=38s) Some examples of these specialized tasks are writing any kind of copy, emails, blog articles, product descriptions. It could be scrubbing fake emails from a list, extracting or normalizing data, translating, paraphrasing, rewriting, qualifying a sales lead, ranking priority of support issues, detecting fraud, or flagging inappropriate content.

[0:59](https://www.youtube.com/watch?v=LksXn4CLC0g&t=59s) These are very common tasks that businesses do every day, and something they have in common is that traditional programming or rule-based approaches do not work well for them. But large language models are great at them.

[1:12](https://www.youtube.com/watch?v=LksXn4CLC0g&t=72s) They perform them easily, and they can capture the nuance in the text that you're working with. So why should we fine-tune? I mean, prompt engineering is great, right? You can do almost all of these things with a prompt.

[1:23](https://www.youtube.com/watch?v=LksXn4CLC0g&t=83s) Well, I'll tell you. Fine-tuning is awesome. It's faster and cheaper because you can train a lighter model to match the quality of what you were doing with a prompt.

[1:33](https://www.youtube.com/watch?v=LksXn4CLC0g&t=93s) It reduces the size of your prompts, allowing for longer completions. Training examples allow you to cover edge cases and collaborate better as a team,

[1:42](https://www.youtube.com/watch?v=LksXn4CLC0g&t=102s) and it's naturally resistant to prompt injection attacks. So let's dive into some of these. How much faster is it really? Well, if you take GPT-4 and its response time per token, it's about a hundred and ninety-six milliseconds, give or take, from the OpenAI API.

[2:00](https://www.youtube.com/watch?v=LksXn4CLC0g&t=120s) On the same API, GPT-3.5 is seventy-three milliseconds. That's three times faster.

[2:07](https://www.youtube.com/watch?v=LksXn4CLC0g&t=127s) How much cheaper is it? Well, taking an example with GPT-4 versus GPT-3.5 fine-tuned, you can save eighty-eight point six percent. Well, then how much shorter do the prompts actually get?

[2:19](https://www.youtube.com/watch?v=LksXn4CLC0g&t=139s) Well, I'll give you one example because it's gonna vary depending on your prompt. But here's what a typical engineered prompt might look like. It has some instructions, um, saying that, you know, write a blog post on this topic, how to write it, what tone to use, what to do, what not to do.

[2:36](https://www.youtube.com/watch?v=LksXn4CLC0g&t=156s) Well, with a fine-tune model, it learns how we write, so we don't need all of those instructions. It learns from our training examples. So we're just giving it the one thing that's unique about this prompt versus another prompt, which is the topic that we want to write on.

[2:51](https://www.youtube.com/watch?v=LksXn4CLC0g&t=171s) And in this very conservative example, it's ninety percent shorter.

[2:57](https://www.youtube.com/watch?v=LksXn4CLC0g&t=177s) Now let's talk about collaborating as a team, right? Because none of us work in a vacuum. We work with other people. Imagine a GitHub repo. You have one file.

[3:06](https://www.youtube.com/watch?v=LksXn4CLC0g&t=186s) Your whole code base is just one file. That's like your epic prompt. Well, with fine-tuning, now you can have multiple files like we're used to, where developers can work on this section of code or that section of code.

[3:17](https://www.youtube.com/watch?v=LksXn4CLC0g&t=197s) But we're not talking about code, we're talking about training examples. So your training data is this layer that your team can work on and add to and edit and improve, and then that feeds into the fine-tune model.

[3:31](https://www.youtube.com/watch?v=LksXn4CLC0g&t=211s) So the main point is if you can get equal or better output, why wouldn't you fine-tune a model? Now, fine-tuning is kind of a dev job right now, okay?

[3:42](https://www.youtube.com/watch?v=LksXn4CLC0g&t=222s) Let's be real. If you go online and you look up how to do fine-tuning, you're gonna find articles that talk about how to spin up GPU servers for training and inference, and you gotta format your data with these ad hoc Python scripts and configure these parameters and then make API calls.

[3:57](https://www.youtube.com/watch?v=LksXn4CLC0g&t=237s) It just looks like a dev job. But if you really break it down, why can't we just automate all of that with a user interface? Is that possible?

[4:08](https://www.youtube.com/watch?v=LksXn4CLC0g&t=248s) It is possible, and the bar is lower than most people think to get started doing this. If you can get twenty examples of what you want your fine-tune model to do, you can fine-tune a model.

[4:19](https://www.youtube.com/watch?v=LksXn4CLC0g&t=259s) This is not traditional machine learning, where you need thousands of examples to get started, and the dataset is this impossible barrier to get past. No, this is something that you could handwrite these if you want to.

[4:32](https://www.youtube.com/watch?v=LksXn4CLC0g&t=272s) One way to think about this is as an extension to few-shot learning. Let's say you can have five examples of what you want a model to do in your prompt.

[4:41](https://www.youtube.com/watch?v=LksXn4CLC0g&t=281s) Well, with fine-tuning, your training example dataset can be as long as you want. So instead of five examples, you can now have twenty or a hundred. So it seems intuitive that with more examples, the model would be able to do closer to what we want it to do.

[4:56](https://www.youtube.com/watch?v=LksXn4CLC0g&t=296s) So here's what I propose for a dev life cycle for large language models. We start with prompt engineering. Prompt engineering is a powerful tool. It allows us to create a prototype, to validate the concept, and we can also use it to create our initial datasets for fine-tuning.

[5:12](https://www.youtube.com/watch?v=LksXn4CLC0g&t=312s) Once we have those datasets, we should fine-tune a model, and we should evaluate it to make sure that it actually is better than the prompt engineered version. And then we can test which models we can get to perform at the same level.

[5:25](https://www.youtube.com/watch?v=LksXn4CLC0g&t=325s) Then the fine-tune model can go into production. And from production, we can capture feedback from our users, and we can log the examples. And with those examples, we can continuously improve our fine-tune model because now all of a sudden we have, um, the real examples that we can add back into our dataset.

[5:42](https://www.youtube.com/watch?v=LksXn4CLC0g&t=342s) So in terms of roles, I think that there's a huge opportunity for people to get into prompt engineering and fine-tuning who are not developers. Yes, if you're a developer, you can fine-tune.

[5:52](https://www.youtube.com/watch?v=LksXn4CLC0g&t=352s) Absolutely. But you shouldn't have to be the only person that can fine-tune. I'm a co-founder at Entry Point, and we have built the modern tooling to make this easy.

[6:02](https://www.youtube.com/watch?v=LksXn4CLC0g&t=362s) Let's take a look at how it works. Here we are on the dashboard, and I'm gonna open the Press Release Writer project. Let's take a look at my 20 examples.

[6:11](https://www.youtube.com/watch?v=LksXn4CLC0g&t=371s) Um, the way I created these 20 examples for a press release generator was I went online and I found 20 press releases that looked really good. They came from blog articles about the best press releases that you can write.

[6:24](https://www.youtube.com/watch?v=LksXn4CLC0g&t=384s) However, I didn't have input data, so my dataset was incomplete, but I used ChatGPT-4 to take the press release and then write a list of facts that would be needed to actually have a professional writer write such a press release.

[6:40](https://www.youtube.com/watch?v=LksXn4CLC0g&t=400s) You know, large language models aren't great at facts, so providing it the facts as the input makes sense to me, that I want to give it a list of facts and then have it write something that's really polished that would be a really good first draft of a press release.

[6:52](https://www.youtube.com/watch?v=LksXn4CLC0g&t=412s) With this user interface, I have a lot of visibility into the data that I'm actually putting into my fine-tuned model, which I think is really important. And the way this works is that we have a structured data approach.

[7:03](https://www.youtube.com/watch?v=LksXn4CLC0g&t=423s) So when you import, like, a CSV into Entry Point, each column becomes a field. Here I have the facts, and here I have the press release. And these fields you can use in a template.

[7:15](https://www.youtube.com/watch?v=LksXn4CLC0g&t=435s) Just like you were writing a mass email and you wanted to insert somebody's first name or personalize the emails with information about a contact record, you can use references to these fields with the Handlebars templating language.

[7:28](https://www.youtube.com/watch?v=LksXn4CLC0g&t=448s) Um, it provides a really intuitive way to easily format your output, your input. And GPT-3.5 Turbo, when you fine-tune it, you can actually use the system prompt, uh, which is where you can include instructions as well, which creates this really interesting hybrid between prompt engineering and fine-tuning, where you can have a small dataset for fine-tuning, but you

[7:48](https://www.youtube.com/watch?v=LksXn4CLC0g&t=468s) can also give it some instructions to help. Once we have a dataset like this, we can go and we can go to our fine-tunes, press the Add button, um, select the model, the platform, because this is cross-platform, and then we count your tokens and estimate your cost for you.

[8:04](https://www.youtube.com/watch?v=LksXn4CLC0g&t=484s) This is gonna be a whole dollar, so [inhales]

[8:07](https://www.youtube.com/watch?v=LksXn4CLC0g&t=487s) hold on tight. Press Start, and that will get started. But I have some here that are already trained, um, so let's go into one and use Entry Point Playground and see if we can actually generate a press release with our fine-tune model.

[8:21](https://www.youtube.com/watch?v=LksXn4CLC0g&t=501s) The list of facts here I actually wrote about, um, the AI Engineer Summit, and we'll see if we can make a press release for the AI Engineer Summit.

[8:30](https://www.youtube.com/watch?v=LksXn4CLC0g&t=510s) Let's go. All right. So this fine-tune model created a title here, and it made it look like a press release. What I found to be a really cool workflow is to actually create a list of facts and then generate an article, read the article, and then get ideas from it, and go back to my list of facts

[8:49](https://www.youtube.com/watch?v=LksXn4CLC0g&t=529s) and refine those, and then that actually becomes an iterative process to get really cool results. So I really enjoy fine-tuning. It takes a lot of the boilerplate out of the prompt, and you can just focus on what's important for the results you want, um, and the rest is taken care of by your training data.

[9:04](https://www.youtube.com/watch?v=LksXn4CLC0g&t=544s) Entry Point has a lot of other cool features, like data synthesis and tools to compare the performance of your fine-tune models. Unfortunately, we don't have time to go into all of that today, but I hope you will check it out.

[9:15](https://www.youtube.com/watch?v=LksXn4CLC0g&t=555s) It's entrypointai.com, and it was a pleasure speaking to you. [upbeat music]
