About
The engineer
behind the systems.
The hard part of AI isn't getting a model to answer once. It's getting it to answer thousands of times, with citations you can check, inside a latency budget, and failing gracefully when the input is hostile.
I'm an AI engineer focused on the distance between a compelling prototype and a system a team can operate. That distance is mostly retrieval quality, agent constraints, evaluation and infrastructure: the unglamorous work that decides whether AI ships.
By day I lead AI engineering at Baseel, with a team of 12. I also founded Tuathra, an AI engineering studio, to build focused production systems for clients. The same engineering standards, applied to problems outside a single company.
I hold an MSc in Data Science and Analytics from MTU Cork, have a paper published with Elsevier, and am a McKinsey Forward fellow. Over 5.5 years I've worked on more than fifteen production projects.
How I think
Build systems,
not demos.
- 01
Context before cleverness
A model is only as good as the evidence you put in front of it. Retrieval, chunking and grounding come first; prompt craft comes last.
In practice: Retrieval quality is measured on its own, before anything is asked of the model.
- 02
Agents need brakes
Autonomy without constraints is a liability. Allow-lists, step budgets, stopping conditions and reversible actions are design work, not afterthoughts.
In practice: Every agent action has a known blast radius, a fallback, and a record of why it ran.
- 03
If it isn't measured, it isn't shipped
Evaluation datasets and traces decide what goes live, not impressions from a demo.
In practice: Model, prompt and routing changes run against a golden set before they reach users.
- 04
Infrastructure is the product
Latency, cost, retries and observability are what separate a prototype from a system.
In practice: Rate limits, timeouts and oversized inputs are designed for, so failure is graceful.
Background
- Baseel
- Lead AI Engineer · team of 12
- Tuathra
- Founder
- MTU Cork
- MSc in Data Science and Analytics
- Elsevier
- Published paper
- McKinsey
- McKinsey Forward
- Experience
- 5.5 years · 15+ projects
Disciplines
- Generative AI & LLMsModel routing · structured outputs · guardrails
- Retrieval (RAG)Qdrant · hybrid search · citations
- AI AgentsLangGraph · tool use · state machines
- Multi-Agent SystemsCrewAI · delegation · long-running workflows
- LLM EvaluationTraces · datasets · regression suites
- AI InfrastructureDocker · Kubernetes · AWS · MLOps
Toolchain
Python · FastAPI · Django · LangGraph · Qdrant · Next.js · Docker · Kubernetes · AWS · PostgreSQL · Redis · OpenTelemetry · CrewAI
Tuathra studio Download CV (PDF)Next step