← All speakers

Bio, Work & Ideas

Vikash Agrawal

Conference affiliation: Amazon Web Services (AWS) · 2025

Vikash Agrawal is a software development manager at Amazon Web Services who leads Amazon CloudWatch initiatives after working on Amazon Q Developer. He focuses on making AI-assisted development practical across testing, deployment, observability, and production troubleshooting.

Earlier in his career, Agrawal contributed to Mozilla and the World Wide Web Consortium through Google Summer of Code and held engineering leadership roles at Housing.com and Dream11. He later worked as an Amazon front-end engineer, participating in a W3C workshop on web games, before shifting toward developer tools, generative AI, and cloud operations.

  • AI throughout the software development life cycle. Agrawal applies coding assistants to planning, implementation, testing, documentation, deployment, and maintenance. A demonstration with AWS developer advocate Linda Haviv follows a FastAPI-based 2048 game through generated unit tests, context-aware bug fixes, API documentation, and Lambda deployment.
  • Production-first AI development. Agrawal emphasizes infrastructure planning, code review, and human accountability. When deployed code fails, CloudWatch logs and AI-assisted investigation help engineers trace operational symptoms to application dependencies and configuration problems.
  • Observability for production AI agents. With Elizabeth Fuentes Leone, he co-created an Amazon Bedrock AgentCore course covering agent runtime, deployment, persistent memory, and monitoring.

Read the topics behind these talks

1 conference talk

Key ideas

Scroll to read ↓

A Python game moves from a CLI prompt to tests, documentation, pull-request review, and a Lambda failure that brings infrastructure back into planning.

  • Where does AI fit in the development cycle?
    0:17 ↗
  • Turn the 2048 rules into a build plan
    2:14 ↗
  • Correct the workspace and inspect the scaffold
    4:01 ↗
  • Test state transitions, then repair an error
    5:26 ↗
  • Refresh documentation after the code changes
    6:55 ↗
  • Use an issue as the development request
    7:49 ↗
  • Investigate the gap between local and deployed behavior
    9:08 ↗
  • Bring infrastructure and prompts into the initial plan
    10:47 ↗

References