Allianz
Regulated enterprise AI
Copilot Studio assistants, voice agents, and AI-driven contact-centre workflows with GDPR, redaction, and audit requirements.
Contract Lead AI Engineer · London · UK remote
Production AI.
Built to ship.
I join enterprise product and engineering teams to design, build, and integrate agentic systems, LLM applications, and AI-enabled workflows — hands-on from architecture and governance through deployment and handover.
- 15+
- Production delivery
- years
- 4
- Agentic AI focus
- years
- 3
- Platforms live
- AIOS · AIBO · ABC
- 115
- Multi-agent workspace
- defined roles
Contract focus
Hands-on leadership for production AI.
I work in the useful middle between strategy and implementation: shaping the system with senior stakeholders, then building it with the engineering team.
Agentic systems
Tool-using agents, multi-agent workflows, MCP integrations, memory, evaluation, observability, and recoverable fallbacks.
Enterprise AI integration
LLM applications connected to real data, APIs, controls, and business workflows — secure, auditable, and reviewable.
AI-enabled engineering
Claude Code and coding-agent delivery systems that improve planning, implementation, review, testing, and deployment.
Selected enterprise delivery
A track record beyond the lab.
Fifteen years across regulated organisations, global platforms, product teams, and data delivery — with AI as the current chapter, not the first one.
RWS
AI workflow integration
AI translation workflows integrated into an established language-services platform across cross-region engineering teams.
Sainsbury’s
Data platform delivery
A CI/CD data pipeline using Snowflake, Jenkins, and Airflow to support marketing analytics at enterprise scale.
How I work
One production loop, then scale.
Fast enough to create evidence in weeks; disciplined enough for enterprise security, governance, and ownership.
- 01Frame the outcome
Define the user, workflow, data, controls, and measurable production result before choosing the model or framework.
- 02Ship a vertical slice
Build one thin end-to-end path through the real stack, including evaluation, observability, and a human review gate.
- 03Prove it under load
Use real runs to improve quality, latency, cost, security, and failure recovery — not a polished demo script.
- 04Transfer ownership
Leave the team with production code, tests, architecture decisions, operating measures, and a practical runbook.
Hiring for an AI programme that must ship?
Send the outcome, current stack, team shape, location, duration, and IR35 status. I reply within one working day.