AI Systems & Evaluation Engineer.
I design and evaluate AI systems, investigate model behavior and failures, and build reliable agentic workflows.
My background spans structural engineering, numerical software and enterprise systems. Today, I bring that experience to AI evaluation, interpretability and system reliability.
Areas of work
- AI evaluation & reliability
- Interpretability & model analysis
- Agentic systems & workflows
- Verification & governance controls
- Scientific computing
Links
Latest work
Research
Same Dice, Different Geometry
An investigation of how identical Dice scores can conceal different geometric structure and connectivity in segmentation, using controlled error families on expert retinal vessel annotations.
Engineering
Agent Reliability & Evaluation
A process for accepting work from AI coding agents: bounded missions, deterministic checks, seeded contradictions and a separate evaluator agent that can veto. Its first system mission was rejected twice.
Publication
PDE-Refined Local Fractal Dimension Prior Conditioning and Topology-Aware Refinement for Retinal Vessel Segmentation with a Swin-UNet-Style Backbone
Lucian Alexandru Murgu and Tudor Barbu. Applied Sciences 16(11), 5559. Retinal vessel segmentation with a fractal-dimension prior and topology-aware refinement.
Publication detailsDOI 10.3390/app16115559 (opens external site)
Interactive explainer
Right Prediction, Wrong Reason
An interactive, synthetic example of a model that is accurate for a fragile reason: swap a shortcut, compare two exact Shapley explanations and test a distribution shift.