Pratik Desai

I’m a CS&E student specializing in AI/ML, focused on building production-ready Generative AI systems—especially RAG pipelines, LLM orchestration, and enterprise AI applications.

At NeuLeap, I architected an AI-powered enterprise database profiling platform using LangGraph autonomous agent pipelines and LangChain with multi-LLM inference. I also engineered semantic column clustering with HuggingFace embeddings and ChromaDB to automate schema documentation, including cross-table relationship discovery and ERD generation.

I build robust, compliant solutions using agentic and evaluation-driven practices—deploying multi-service microservices with Docker Compose and connecting via dual-database persistence (PostgreSQL, DuckDB, ChromaDB) while keeping enterprise connectors read-only to avoid data exfiltration risk.

Across my projects (ATLAS Copilot, Unified Data Profiler, and Adaptive RAG), I deliver end-to-end systems with hybrid retrieval, reranking, tool calling, and stateful orchestration using LangGraph/Semantic Kernel. I’m excited by roles where I can keep shipping reliable LLM products, improve quality using monitoring and hallucination detection, and scale them with LLMOps and evaluation.