Ronak Ajwani

I’m an AI Engineer and Computer Engineering graduate building LLM applications with RAG pipelines, multi-agent systems, and evaluation harnesses that prove the system actually works. I don’t ship “maybe”—I build golden datasets, faithfulness scoring, and guardrails from the start.

In my Tata Communications AICoE internship, I built an agentic code-review pipeline solo, orchestrating specialized agents for security, bugs, documentation, and refactoring. I mapped every finding to OWASP Top 10 and SOC 2, designed a 3-layer false-positive filter that cut error rates from 30% to 5%, and ran a model bake-off across 531 OWASP benchmark cases to choose the best-performing model on evidence.

I bring the same rigor to my projects: NotebookRAG delivers document QA with hybrid retrieval (dense + BM25) plus an evaluation harness, including LLM-as-Judge scoring and an abstention gate that prevents hallucinated answers. I also shipped QuantScope on Azure/Vercel for real-time financial analytics and built Project Abyssa, a sandbox-validated code-repair agent that only surfaces fixes once the tests pass—then earned first place in the Agentic AI Track at Syrus Hackathon (170+ teams).