I’m Nikhil, an AI & ML engineer building production ML systems — RAG pipelines with retrieval evaluation baked in, CI quality gates that catch regressions before they ship, observability that traces a request from prompt to response.

I have a background in both the business and technical sides of ML. Three years at Microsoft as an ML & DS Consultant on large Azure delivery portfolios taught me to care as much about latency SLAs, cost-per-request, and failure traceability as benchmark numbers. I completed my M.S. in Computer Science at Rice University in 2024, where I also collaborated with NASA through Rice’s D2K Lab on spacecraft instance segmentation for onboard embedded hardware.

This site is where I write up what I build, the technical decisions behind it, and what I’d do differently next time.

---

📄 Publications

SWiM: A New Dataset and Performance Benchmark for Real-time Spacecraft Segmentation in Onboard Computers · arXiv:2507.10775 Co-lead author · D2K Lab / NASA · 2024

64K annotated images and the first standardized benchmark for real-time instance segmentation on embedded orbital hardware.

---