Software engineer, ML engineer, UI/UX designer, web developer and app developer.
I build agents that act, and the guardrails that stop them when they shouldn’t. Mostly self-taught. Always shipping.
Third-year CS student at SRM building backend and AI systems: tool-calling LLM agents, hybrid retrieval pipelines, and the evaluation and safety layers that keep them reliable. Open-source maintainer, GSSoC mentor, and open to internships and collabs.
A proactive safety app that replaces the passive SOS button.
Dynamic ML threat mapping, safe routing that avoids high-risk zones, and on-device audio threat detection with YAMNet that raises an alert and counts down to an automatic SOS.
Turns vague complaints into verified, one-tap device fixes.
Hybrid BM25 + dense retrieval grounds a multi-stage LLM pipeline across Gemini and Mistral, and every plan comes with verified Galaxy Settings deep links. A semantic cache answers in under 300 ms.
Your SRM attendance, marks and timetable, minus the portal.
An installable app that signs into the academia portal for you: a bunk predictor that says exactly how many classes you can still miss, a leave planner, a grade forecast, and a timetable that runs on day orders.
Spots revenue at risk and runs a bounded recovery workflow.
A deterministic taxonomy picked the right action on 9 of 11 causes, the LLM on 3. So the model only sees unseen error codes, behind a guardrail it can’t override: budget, contact caps, quiet hours.
Internships, collabs, hackathon teammates, or a 2 a.m. debugging rant: my inbox is open. Pick a reason, tear off the receipt, and it opens in your own mail app. I usually reply within a day.
nikhilbalamurugan@gmail.com
Turning caffeine into code.
