Jeet Majumder

AI Engineer

ViniyogOne • Temps plein

Apr 2026 - Present • 5 mos

Started as an AI Engineer Intern architecting enterprise Retrieval-Augmented Generation (RAG) pipelines with hybrid retrieval, semantic chunking, and source-grounding to minimize hallucination rates, subsequently promoted to full-time AI Engineer to lead the design and deployment of an Agentic AI Financial Coach. Engineered stateful multi-agent workflows using LangGraph, Google ADK, and DeepAgents, implementing autonomous tool calling, MCP integrations, persistent conversation memory, and Human-in-the-Loop (HITL) review patterns. Built automated evaluation pipelines, Pydantic structured output validation, and production-grade async REST microservices using FastAPI, Docker, and CI/CD pipelines to ensure low-latency, policy-compliant execution in financial services.

Junior Data Scientist

Databae • Temps plein

Mar 2025 - Sep 2025 • 6 mos

Joined as a Data Science Intern researching computer vision and object detection architectures, advancing to full-time Junior Data Scientist to engineer and deploy real-time surveillance and edge AI systems. Developed a multi-camera CCTV person tracking and re-identification pipeline utilizing body and face embeddings to resolve complex visual occlusions across distributed video streams. Built an automated door-level entry/exit monitoring pipeline achieving sub-second detection latency to eliminate expensive third-party vision API reliance, while training, pruning, and benchmarking custom YOLO-based models for real-time workplace activity monitoring and hazard detection.