At Multiverse Computing, I developed a production-grade multi-agent chatbot with LangGraph, routing requests to specialized RAG, pricing, and scheduling agents. It also included LLM-powered multilingual translation.
I designed a modular RAG framework with configurable data and API connectors, multiple vector storage backends, reranking-based retrieval, long-term memory, and guardrails. I also architected a GraphRAG assistant that turns unstructured business data into Neo4j knowledge graphs.
Through the Mitacs Globalink Program, I applied quantum and classical probabilistic machine learning models to brain activity patterns, identifying neurological biomarkers that distinguish autistic from neurotypical brain regions. The models achieved 80% accuracy and 93% recall on real-world medical datasets.
At Orange Group, I developed a crypto wallet Chrome extension with voice biometric authentication, using deep learning embeddings for user verification with over 87% accuracy. At Next Generation IT, I designed a pneumonia detection model and developed Apache Airflow pipelines for automated machine learning workflows.
