Applied Research

Questions before conclusions.

Research-led work where I start with a gap, investigate the problem, test an idea, and turn the useful parts into something that can be built.

Work projects

Research in practice
Hand-drawn illustration for LLM and Agentic AI application evaluation

LLM & Agentic AI Application Evaluation

How do we decide whether an AI application is actually working well?

Research direction

Exploring evaluation approaches for agentic and multi-agent applications, including how evidence can be captured, compared and surfaced for technical and governance decisions.

My contribution

Conceived the problem space, researched the gap, shaped the approach and proof of concept, and drove implementation into reusable evaluation and governance capabilities.

Personal projects

Independent explorations
Hand-drawn illustration for Generic Adaptive RAG Framework

Generic Adaptive RAG Framework

Can a retrieval system recommend its own configuration from the characteristics of the documents it needs to answer?

A framework concept for analyzing document collections and recommending retrieval configurations and hyperparameters, with the aim of reducing manual RAG tuning.

Hand-drawn illustration for Hyperparameter Optimization Framework

Hyperparameter Optimization Framework

How can hyperparameter search make better use of a limited training-time budget?

A neural-network training strategy for time-constrained hyperparameter optimization, initially validated across multiple CNN training scenarios.