Akshay K Upadhyay

Desktop app that marks attendance and counts visitors from a live camera: YOLOv11 person detection, InsightFace face embeddings and centroid tracking with IN/OUT line crossing, in a threaded pipeline that stays real-time on ordinary laptops.

Team project (8 people) for the IIT Madras Software Engineering course. I built the backend and much of the Vue frontend: a Gemini RAG chatbot that answers from company policies and sends unanswered questions to HR, AI resume screening and job description writing, leave management, and dashboards for HR, managers and employees.

Predicted customer purchase value from behavioural data. 150+ tracked experiments in encoding, feature engineering and model selection lifted the validation score from below the 0.45 baseline to 0.73.

Multi-agent consensus system utilizing multiple models (Claude, Gemini, etc.) to evaluate and judge complex conversational data.

Built Retrieval-Augmented Generation (RAG) pipelines using LangChain and semantic retrieval workflows. Implemented document ingestion pipelines, embedding workflows, vector search systems, and contextual AI response generation.

Built a fullstack AI-powered analytics platform using FastAPI backend services and React frontend architecture. Developed LangChain workflows for natural language querying, reasoning pipelines, tool execution, and automated report generation.

AI-powered tool to automatically generate complete presentation slides and talking points based on user prompts.

Robust platform connecting users with local service professionals through direct hiring or bidding, featuring an admin dashboard and professional verification.