About Soham Chaudhari

I am currently an AI Engineer at FLIP, a platform that helps people find card offers, earn more rewards and redeem points well. Before that I worked across fintech, legal-tech and consumer products, always on the applied side of AI: taking a model or an idea and making it reliable, fast and affordable enough to ship.

Most of my work involves agentic systems: LLMs that plan, call tools, read documents and act. I also build retrieval pipelines, voice assistants and vision models, and I care about the unglamorous parts that decide whether an AI feature survives contact with real users: evaluations, latency, cost and failure handling.

Agentic AI and multi-agent systems with LangGraph and LangChain. LLM orchestration, tool calling and the Model Context Protocol (MCP). Retrieval-augmented generation over documents and video. Voice AI. Computer vision, including models that run entirely in the browser with ONNX Runtime Web. Backends in Python and FastAPI.

MOSAIC searches and explains long videos with multimodal retrieval. Stock AI is an LLM agent that answers questions about stocks with live market data and also ships as an MCP server. GPTs is a decoder-only transformer written from scratch in PyTorch. VISION AI restores low-quality images and video. The face detection and MNIST demos run neural networks directly in your browser.

I studied at Vivekanand Education Society's Institute of Technology (VESIT) in Mumbai. Along the way I placed in several hackathons, including first place at Syrus'26 and second place at IIT Ropar's Medino's x Advitiya'25.