I build systems that know when not to trust themselves.
Full-stack and AI engineer, currently orbiting Dublin. I ship production software, then spend just as long proving it's actually reliable before I let anyone near it, myself included.
How I actually work, no mission statement fluff.
01 · Build first, trust never
Shipping is the easy part. Every project gets a real evaluation layer, not a gut feeling, that tells me exactly how much to trust its own output before anyone else has to.
02 · Test what should break
I build adversarial cases specifically to make my own systems fail, so the failures show up on my desk, not on a stranger's screen at 2am.
System readout.
Things I've actually shipped.
No filler projects, no tutorials-in-disguise. Everything below is live, tested, and real people have used it.
A production agentic AI platform built end to end, alone. LangChain orchestration, a RAG pipeline over ChromaDB, and a three-tier confidence-gating evaluation layer that explicitly declines to answer when retrieved evidence falls below a calibrated threshold, rather than guessing.
A Linux automation CLI that replaced a manual, multi-step provisioning process with a single command. Hermetic subprocess mocking means the test suite runs the same way every time, on any machine, in any orbit.
An end-to-end insurance management application, tested against real validated record volume, with zero defects found across a 200-record audit. Boring in the best way.
The actual spec sheet.
Where I've actually flown.
- Reduced client manual processing effort by 60% by designing and deploying AI-assisted automation into production workflows.
- Served 200 active users with zero downtime by architecting fault-tolerant systems with automated recovery.
- Scaled a client system to 10x its original data volume without added infrastructure, by fixing the actual bottleneck instead of buying a bigger engine.
- Improved client data accuracy by 40% through automated validation processes.
- Achieved zero production incidents through automated health-check monitoring and systematic root-cause analysis.
- Delivered analytical solutions to 6 stakeholder groups, translating datasets into clear, actionable recommendations.