At minto.ai, I deployed an autoencoder-based anomaly detection model that scores steel-plant motors live from daily readings. I also built an alerting engine that uses Grey Wolf Optimizer and rule-based logic to notify plant users.
I built the SpiderAI⢠platform, where an LLM turns uploaded maintenance and diagnostic files into wikis using a domain ontology. After human approval, the platform builds a Neo4j knowledge graph that an Ask agent uses to answer questions.
I built ResearchMind, a multi-document agentic RAG application with a React frontend and FastAPI backend. It can fall back to DuckDuckGo when confidence in the uploaded documents is low, and supports authenticated sessions with isolated FAISS indices.
My other projects include a full-stack e-commerce platform and a BERT-based sentiment classifier. I also built a deepfake video detection system using EfficientNet-B4 and deployed it as a Flask REST API with a JavaScript frontend.

