Soham Chaudhari

Engineered an 11-node LangGraph orchestration pipeline and a 4-step hybrid retrieval engine to power an AI credit card recommendation system over 9.7K+ offer entries. Built PostgreSQL semantic memory and production APIs, optimizing memory injection to top 1–2 items and reducing token usage by 60–70%, and deployed the backend on AWS EC2 with Docker Compose and rate limiting at 100 req/min.

Built an AI credit card recommendation platform using a LangGraph workflow and hybrid retrieval (semantic search, web Search fallback) over 9.7K+ offers. Developed PostgreSQL-based semantic memory with session isolation and feedback APIs, cutting token usage by 60–70%. Deployed on EC2 with Docker, connection pooling, health checks, rate limiting, and production-grade reliability.

Deployed end-to-end Legal Operations AI agents using LLMs, reducing manual document review time by 50%. Improved LLM contextual accuracy by 31% via prompt engineering and architected RAG pipelines with Mistral AI, LangChain, and Pinecone to increase retrieval accuracy by 60% and eliminate hallucinations.

Implemented and deployed a recommendation system achieving 95% accuracy, led development of an AI-based visa training system using TTS, STT, and LLMs, and engineered REST APIs (Node.js, Express.js) that reduced response times by 40%.

Built a financial chatbot with 95% accuracy that reduced response time by 60%, evaluated 10-K/10-Q datasets for major tech firms, and developed an NLP pipeline to extract financial KPIs for faster forecasting.

Grade: 8.11 CGPA

Bachelor of Engineering in Computer Engineering with a CGPA of 8.11.