Chirag Agrawal

At Amazon, I built and productionized an ML-powered Alexa podcast search system using BM25 and LambdaMART, reducing search overrides by 37%.

I analyzed retrieval failures and fine-tuned ranking strategies to reduce sparse-ranker errors to 4%. I also extended the stack with LLM-based retrieval, ranking, tool-calling, MCP, agent skills, LLM-as-a-Judge, and A/B testing.

At Sprinklr, I built a Google Docs/Notion-style rich text editor with TipTap, ProseMirror, React, and TypeScript across 37 editor plugins. I also created an agentic editor-development tool to automate repetitive modifications and accelerate plugin work.

I research quantitative alpha strategies at Trexquant and have produced 20+ alphas with IR above 0.10. My projects include a multimodal phishing detector reaching 97.88% accuracy and self-driving agents achieving 95%+ lap completion on unseen tracks.