Led the Image Processing and Machine Learning team to design and implement the AI-driven computer vision automation pipeline for R3Bin, a smart waste-segregation dustbin.

Built and deployed real-time waste detection pipelines, seamlessly integrated with robotic actuation for automated waste sorting

Built the complete Retrieval-Augmented Generation (RAG) pipeline using LangChain, Ollama (Mistral-7B), and vector embeddings for secure offline document querying.

Developed Python TCP socket communication enabling seamless client-server query and response exchange over a LAN

Built an AI agent for developer recommendation using a RAG-based system that parses developer skills, availability, and internal ratings to match client prompts with suitable engineers.

Designed and implemented the RAG pipeline and scoring system, incorporating GitHub commit activity, Waka-Time usage data, and Copilot-based code review scores to rank candidates.

Built a full-stack on-chain AI agent auth protocol on Base Sepolia using ERC-8004, Biscuit WASM, TLSNotary,and ERC-4337; shipped a 12-phase live demo with SSE streaming, O(1) cascade revocation, and a React console for multi-agent orchestration.

Designed an ML-driven trust scoring engine for autonomous AI agents using a 3-signal model (scope adherence, task completion, time-decayed quality) to gate on-chain delegation across four tiers, with defenses against sybil attacks, burst farming, and history padding.

Designed an algorithm to autonomously align the bot to it's counterpart and the basketball hoop

Designed the vision and path planning algorithm for the Auotonomous bot using various ML and DL algorithms.

Won the first prize in the mathworks modelling award 2026.