At Brij Udyog Vihar, I built and deployed a CRM platform used by 15+ staff daily to manage property listings, leads, and client interactions. I implemented live collaboration with WebSockets and used Redis-based distributed locking to prevent concurrent edit conflicts.
At IIT Delhi, I developed and fine-tuned computer vision algorithms to remove rain streaks from real-world images and videos. I used gradient-based rain detection and directional filtering with OpenCV and Python, and prepared datasets, trained models, and validated results on benchmark datasets.
On my tiny-hnsw project, I implemented an HNSW engine from scratch and measured sub-2ms P50 search latency with up to 89% recall@10 on 10K vectors. I’m currently integrating Qdrant with LanceDB for a hybrid search pipeline, and I also built LMAX-rs, a Rust order matching engine that sustained 1.0–1.3M messages per second in stress testing.
