Vikash Agrawal

Vikash Agrawal is an AWS software development manager who has worked on Amazon Q Developer and CloudWatch initiatives. His public demonstrations show AI-assisted development extending beyond code generation into reviewing changes, repairing tests, documenting applications, and investigating deployment failures. His undated GitNation biography describes earlier involvement in developing Amazon Q Developer and leadership in the CloudWatch team; his Analytics Vidhya instructor biography specifies CloudWatch Logs and AI initiatives.[2][4][6]

In Supercharging Your Developer Workflow with Amazon Q Developer, published March 27, 2025, Agrawal and fellow AWS presenter Abeetha Bala demonstrated building a 2048 game. Agrawal began with Python boilerplate and prompted Q Developer to create a front end and back end. He inspected the agent’s plan and generated changes, noting that it recognized the FastAPI application without being told and preserved existing configuration and logging. He reported that building the application took around 13 minutes, while acknowledging that the presentation had been edited.[4]

A working game was only an intermediate step. Agrawal asked, “Is it production ready? Probably not.” He reviewed generated tests before accepting them, worked through failures affecting both tests and source code, and removed a duplicated patch introduced while pasting changes. He reported 24 passing tests, then requested negative tests specifically for the make-move function and reported 25 working tests. He also demonstrated generating a README with installation instructions, API documentation, testing steps, and a data-flow diagram, and accepted a code-review fix for inadequate error handling.[4]

For deployment, Agrawal proposed S3 and CloudFront for the front end, Lambda and API Gateway for the back end, and Lambda layers for dependencies. The joint session then showed a deployment problem investigated through Q Developer and CloudWatch suggestions and hypotheses. Its closing account reported getting the application working through trial and error, but concluded that it was unsuited to serverless architecture and described an EC2 deployment as a future step. Later that year, Latent.Space’s AI Engineer World’s Fair recap identified Agrawal as an Amazon Q Developer presenter from the team.[4][1]

Agrawal and Elizabeth Fuentes Leone are listed as instructors for Building AI agents with Amazon Bedrock AgentCore, covering runtime, long-term memory, and agent observability. His announced JSNation US 2026 talk, “When AI Writes Your Code, Who Monitors It?”, addresses unpredictable error patterns, missing instrumentation, and tests that miss consequential failures in AI-generated code. GitNation lists the presentation as upcoming.[6][2]

His GitHub account, ivikash, hosts TabGenius, a Chrome extension documented as grouping browser tabs by their content rather than only their domains. Its README describes Chrome Gemini and local Ollama model options, custom categorization prompts, and an undo function. The documentation says no data is sent to external servers when local models are used, and qualifies results by model quality, tab count, and the chosen model.[3][5]

Agrawal’s earlier engineering leadership includes teams at Housing.com and Dream11, according to his GitNation biography. That biography also reports Google Summer of Code contributions to Mozilla in 2012 and W3C in 2013. His earlier developer-community activity provides concrete historical context for his later public teaching: the official Mozilla Reps archive records membership from January 15, 2012, to December 6, 2013, and organizing roles for MozCamp Manipal on March 20, 2013, and “Mozilla Talk and MDN DocSprint with Ali Spivak” on September 26, 2013.[2][7]