Technical Architect + Applied AI Engineer
I design and build dependable AI and distributed systems where state, reliability, observability, recovery and controlled authority matter.
Evidence driven evaluation for Agent Skills using deterministic analysis, behavioural trials, fail closed gates and verified repair.
A deterministic autonomous decision runtime for discovering work, routing risk and producing replayable, auditable outputs.
Loss bounded context compression for model handoffs, worker evidence and frontier review, preserving decision critical meaning while reducing repeated context.
A research notebook for extracting reusable architecture mechanisms from tools, systems and experiments, then reducing them to what is actually needed.
Risk based model orchestration that keeps high consequence judgement with frontier models while delegating bounded execution with evidence based acceptance.
Runnable Python reference implementations for agent orchestration, reliability, memory, coordination and tool use patterns.
Agentic systems · AI evaluation · RAG and retrieval · Human governed autonomy · Distributed systems · Reliability · Observability
12+ years across production software, regulated data, defence training systems and distributed platforms. Recent work applies those same engineering disciplines to agent orchestration, retrieval, evaluation, guardrails and production AI.