MSc Robotics at the University of Birmingham · Mechatronics engineer.
I build robots, production software, and open-source AI tooling — and I publish the numbers and the bugs, not just the wins.
🌐 pieroscv.com · 💼 LinkedIn
- Robot-qarm-fruit-sorting — Autonomous fruit sorting on a Quanser QArm 4-DOF (Team of 4; my part: camera–arm calibration, detector tuning, voice control, GUI concurrency). 0.0002 mm IK round-trip error, 13-state FSM, offline voice control.
- Autonomous-Navigation-and-3D-Mapping-for-Grass-Cutting-Robots-on-Inclined-Terrains — Slope-aware navigation for a Clearpath Moose in 2,600 lines of raw C — no ROS (Team of 4; my part: LiDAR pipeline, OctoMap-style 3D mapping + ICP, EKF sensor fusion).
- pieroscv — This portfolio: static Astro with zero client-side JavaScript, yet 53 automated tests on every push (full-page visual regression + WCAG 2.2 AA).
Tools that make AI-assisted work actually rigorous:
- dr-joe — a Claude Code skill that grades academic work against a real rubric, with worked examples spanning the grade range.
- dr-pavlov — a thesis examiner that refuses to grade until you hand it the rubric.
- forging-skill — a pre-planning router for AI agents, built test-first on prompts with the TDD evidence published.
- claude-handoff-skill — session closure and relay for parallel AI coding sessions, as an append-only state machine.
- panoptes — a complete gym tracker in one 33 KB HTML file: offline-first, no build step, no dependencies.



