MSc Artificial Intelligence · University of Kent
Building trustworthy AI systems that work in the real world.
My MSc dissertation focuses on Trustworthy Multimodal RAG for Medical Decision-Making: a system that combines retrieval-augmented generation with multimodal inputs (text + imaging) to support clinicians, with explainability and safety at its core.
Core research themes:
-
Trustworthiness: uncertainty quantification, hallucination mitigation, source attribution
-
Multimodality: fusing clinical text with medical imaging in a RAG pipeline
-
Medical AI: responsible deployment, regulatory-aware design (NHS, NICE guidelines, data sovereignty) Alongside the dissertation, I'm building a portfolio of interconnected, production-grade RAG projects:
-
AgenticRAG-Router: An LLM agent that dynamically selects between BM25, dense, and hybrid RRF retrieval strategies, exposed as MCP (Model Context Protocol) tools and benchmarked against static retrieval baselines.
-
MedRAG-Eval: A pipeline-agnostic evaluation framework for RAG trustworthiness in biomedical settings, built around three pillars: retrieval quality (Hit Rate@k, MRR, nDCG), groundedness (faithfulness, hallucination rate, citation accuracy, abstention), and answer faithfulness (correctness, semantic similarity, numeric exact-match). Benchmarked on PubMedQA (
pqa_labeled, 1,000 expert-annotated biomedical QA pairs) and shipped as a pip-installable library, not a Docker service. The two are designed to connect: MedRAG-Eval evaluates the same retrieval pipeline built for AgenticRAG-Router, so the dissertation, the agent, and the evaluation framework reinforce each other as one coherent system rather than disconnected proofs of concept.
Languages
AI / ML
RAG & Agentic AI
Data & Visualisation
Web & Backend
Auth & API Security
Retrieval-Augmented Generation (Agentic Retrieval, Hybrid Search, RRF Fusion)
RAG Evaluation & Trustworthiness (RAGAS, LLM-as-Judge, Groundedness)
AI for Healthcare & Clinical Support
Responsible & Explainable AI
Async Backend Systems (FastAPI + SQLAlchemy 2.0)
Secure API Design (JWT Authentication, Password Hashing)
Open to research collaborations and AI/ML roles, especially RAG engineering roles in regulated sectors.