Professional work
Turning research questions into reliable software
My work connects reinforcement learning, planning and research software. I build experiments that can be audited, communicate what the evidence supports, and turn difficult ideas into tools other people can inspect.
Current roles
Research Software Developer
June 2026–present
Building reproducible multi-agent LLM experiment software for management research, including configuration, structured logging and auditable runs.
Doctoral Researcher
September 2024–present
Researching planning and reinforcement learning for energy-efficient building control, interpretability and reliable learning.
How I work
A reproducible research loop
Frame the decision
Start with the operational question, constraints and evidence needed to make a useful claim.
Build the comparison
Define baselines, controlled changes and evaluation conditions before interpreting results.
Make runs auditable
Use explicit configuration, structured outputs and checks that make experimental runs easier to reproduce.
Explain the evidence
Separate what was measured from what is inferred, then communicate the result at the reader’s level.
Open work
Selected public projects
Applied AI
Production engineering
Associate AI Engineer · CureMD
July 2023–September 2024
- Built an internal LLM and RAG code assistant used across 150+ developers, helping reduce code-review turnaround by more than 25%.
- Shipped RAG support systems handling 500+ daily queries with a reported 92% resolution rate.
- Developed automated code-grading workflows using Llama 3 and programmatic tests.
Teaching and service
Supporting research communities
Teaching Assistant and Lab Demonstrator
Led weekly deep learning and computer vision labs for 30+ MSc Data Science students and graded 200+ assignments per semester.
Volunteer Teaching Assistant
Supported module delivery and student learning beyond contracted duties.
- Oral presentation at AAAI 2026 on deadline-aware, energy-efficient domestic hot-water control.
- Talk at the Tri-University International AI Symposium, 2026.
- Co-organiser of the Coventry–NUST–Monash Suzhou symposium, April 2026.
- Peer reviewer for Elsevier and MDPI journals.
Research
Questions I work on
Energy-aware control
Planning and reinforcement learning for building energy systems under changing prices, deadlines, uncertainty and comfort constraints.
Reliable reinforcement learning
Understanding how numerical details can affect policy optimisation and making experimental claims reproducible and easier to audit.
Interpretable policies
Turning learned control behaviour into compact explanations that people can inspect, compare and reason about.
Research software
Building configurable, structured and repeatable software for AI experiments, including multi-agent LLM studies.
Recognition
Teaching, funding and awards
Tools
Technical toolkit
Languages
- Python
- C++
- MATLAB
Reinforcement learning and control
- Stable Baselines3
- RLlib
- Gymnasium
- MuJoCo
- BOPTEST
- Monte Carlo Tree Search
- SINDy
Machine learning
- PyTorch
- TensorFlow
- scikit-learn
- OpenCV
- Hugging Face
Research infrastructure
- Git
- Docker
- Linux
- HPC and SLURM
- Weights & Biases
- LaTeX
Explore the work
Read the publications, try the interactive demos, or download the verified CV for a concise record.