Real-time CV pipelines, LLM inference at the edge, MLOps on air-gapped infrastructure — I own the problem end to end, from GPU kernel to production dashboard.
Core work
Deep in
Inference optimization → TensorRT · Quantization · Tiled inference · <50ms on constrained hardware
RAG & Agentic systems → Ground-up design, not wrappers. Real retrieval for real problems
Vision at scale → Detection · Tracking · Geospatial analysis · Production deployments
ML Platform → Training → Serving → Monitoring · CI/CD · Drift detection · Rollback
Tools don't matter. Problems do.
M.Tech — AI & Data Science · IIIT Kota / MNIT Jaipur