navneet kumar

I'm a Computer Science student at IIIT Delhi (Class of 2026) and a software engineer who enjoys building complete products not just models or endpoints in isolation, but the full system: databases, APIs, frontend, and the ML/LLM pieces that make a product genuinely useful.

My three main projects reflect that range. agentRelay is a backend service (Node.js, Express, PostgreSQL, MCP) that gives AI coding agents persistent memory of project state — tasks, decisions, failures — so work isn't lost between sessions; it involved designing a retrieval/ranking system over pgvector to compile minimal, relevant context on demand. MerchantScore is a full-stack credit-scoring platform for Indian MSMEs (FastAPI, React) where I engineered financial features, trained XGBoost models, and built a SHAP-based explainability layer with LLM-generated natural-language output. Mirror is a fashion recommendation and virtual try-on platform combining LLM-based feature extraction with pgvector semantic search across a 1,000+ item catalog.

My core technical strengths are backend engineering (Node.js/Express, FastAPI, PostgreSQL, REST API design) and applied AI/ML (LLM integration, prompt engineering, RAG, vector embeddings, semantic search, XGBoost, SHAP). I completed Andrew Ng's Machine Learning Specialization, hold a Codeforces Pupil rating, and have solved 200+ LeetCode and 60+ SQL problems.

I'm most interested in software engineering and AI engineering roles ideally at a company where I can work close to the product, take ownership of systems end-to-end, and keep building at the intersection of solid engineering and applied AI. I'm especially drawn to teams working on developer tools, agentic systems, or AI-powered products with real-world impact.