Jaideep Kundu

I'm a final-year B.Tech CSE student at VIT-AP, and I like building infrastructure rather than one-off features: systems where latency, reliability, and correctness can be measured.

At Bluestock Fintech, I worked on the data layer of a Nifty 100 equity analytics platform. I built idempotent ETL pipelines for financial statements across 92 companies and a 30+ KPI ratio engine served through FastAPI. I also built a reporting pipeline that generates 91 company tearsheets and 10 sector reports per cycle, replacing a manual, analyst-written process.

On the backend side, I wrote an Operational Transformation engine from scratch for CollabDocs, a real-time collaborative editor. I then scaled its sync layer horizontally with Socket.io and Redis Pub/Sub, keeping update propagation under 50ms with no duplicated messages across server instances. In another project, I built a metadata-driven runtime that generates a full CRUD app (UI, REST APIs, and persistence) from JSON config, with zero per-entity code.

For AI engineering, DeepContext Engine is a multi-tenant RAG platform with two-stage retrieval: ANN search over Gemini embeddings, refined by cross-encoder reranking. It handles 7 document formats with page- and slide-level citations. A multi-tier LLM fallback keeps requests alive when a provider goes down.

I try to evaluate as rigorously as I build:

• RTO Shield scores cash-on-delivery return risk with a calibrated LightGBM model on 120K+ orders. It turns those scores into a pricing policy that saves Rs. 53K per 10K orders, with bootstrap confidence intervals behind the gain.

• RiskFrontier is a Monte Carlo VaR/CVaR engine and walk-forward backtester over NSE data. It shows a 31% CVaR gap at the 99% level that a normality assumption hides, and it flags its own survivorship bias.

• MedDiagno tunes decision thresholds per disease to favor recall, because in screening a missed case costs more than a false alarm.

Outside projects, I was a finalist at the Bhopal Health Hackathon with Johns Hopkins University, and I've led hands-on computer vision sessions using OpenCV and MediaPipe.

I'm looking for a 4–6 month internship starting January 2027 in backend/SDE, AI engineering, or data and quant analytics, and I'm open to relocating.

Python · TypeScript · SQL · C++ · FastAPI · Node.js · React · Next.js · PostgreSQL · MongoDB · Redis · LangChain · Pinecone · LightGBM · Docker · AWS