Projects

End-to-end Retrieval-Augmented Generation platform built on Azure. Ingestion pipelines, vector search, hybrid retrieval, and a chat interface — all production-hardened with observability and RBAC.

Reusable Azure ML pipeline framework for training, evaluation, and deployment of ML models. Supports experiment tracking, model registry, and automated rollback on regression.

Automated evaluation harness for LLM outputs — measuring factual accuracy, groundedness, coherence, and latency across model versions. Integrates with CI/CD to gate deployments.

Scalable document search service using dense vector embeddings and BM25 hybrid ranking. Processes millions of documents with sub-100 ms retrieval latency.

Multi-agent orchestration system for complex enterprise workflows. Routes tasks to specialised sub-agents, handles tool use, and maintains conversation state across long-running sessions.

This site — a minimal personal blog and portfolio built with Next.js, Tailwind CSS, and MDX. Statically generated, zero client-side JS on article pages, and deployed to Azure Static Web Apps.