Luke Stedman · Founder & CTO

Bringing applied AI into regulated production.

Founder and engineering leader — 25 years deep in platforms, data infrastructure, and trading systems for regulated and quantitative finance. Today: CTO at a London long/short hedge fund, and founder & CTO of an AI risk platform for the London insurance market.

About me Get in touch

About

I'm a founder and engineering leader. For 25 years I've built platforms, data infrastructure, and trading systems for regulated and quantitative environments — the kind where “mostly works” isn't good enough and every change has to be auditable.

Today I'm CTO at Regents Gate Capital, a London long/short hedge fund, where I built the technology function from zero. I'm also founder & CTO of IRIS — an AI risk platform automating the document-heavy work underwriting analysts do by hand today.

Earlier: J.P. Morgan, Winton, and ClearMacro — data-mesh design, event-driven micro-service frameworks, SRE leadership, and the full arc from research developer to firm-wide platform owner. My focus now is getting frontier AI into regulated production with the guardrails the domain demands.

The Journey

  1. 2001 — 2008

    Foundations — Data & Transformation

    Broadcasting Dataservices · Tideway Systems

    Cut my teeth on XML / XSLT data pipelines and Unix service-discovery models — turning messy feeds into reliable, structured output. Led a UK data team, then wrote application-fingerprinting patterns for enterprise discovery.

  2. 2008 — 2021

    The Winton Era

    Winton · 13 years, five roles

    From research developer to Head of Data Projects. Built Winton's first model-deployment tooling, migrated trading systems to containers ahead of the wave, ran an 8-person SRE team that cut firm-wide MTTR from ~8 hours to under one, and owned US technical strategy across NYC and SF.

  3. 2021 — 2023

    Platform Modernisation

    ClearMacro · J.P. Morgan

    Led a 6-developer rewrite of ClearMacro's signal platform from F# monolith to containerised Python — deploy cycles from weeks to a day. Then at J.P. Morgan's Capital Connect, shipped an event-driven nano-service framework on EKS that the ML team adopted as its integration fabric.

  4. 2023 — Present

    Building the Function — RGC

    Regents Gate Capital · Head of Data Engineering → CTO

    Joined as the firm's first technology hire and built the data and engineering capability from zero. Now CTO, owning the whole stack — investment-data platform, delivery, security, BCP/DR — and driving an LLM-native engineering workflow at ~3× throughput, plus an agentic operator for the signal pipeline under full human governance.

  5. 2025 — Present

    The Founder Era — IRIS

    Instant Risk Intelligence Systems · Founder & CTO

    Founded IRIS alongside the RGC seat: an AI risk platform automating document extraction for London-market underwriters. Eight Rust modules on a per-tenant, event-sourced, AWS-serverless stack, a multi-provider LLM core, and regulated-AI deployment patterns — per-tenant isolation, full audit trail, concentration-risk routing.

Selected Work

IRIS — AI Underwriting Platform

Eight Rust modules on a per-tenant, event-sourced AWS-serverless stack; a multi-provider LLM core feeding a 156-feature cyber-insurance catalogue. Document → per-page XML → guaranteed-schema extraction.

  • Rust
  • AWS Serverless
  • LLM
  • Event Sourcing

Agentic Signal Operator

A PM converses with Claude to propose parameter changes; the agent stages the diff, runs the backtest and forward scenarios, and presents results for approval — full audit trail. The LLM never invents signals.

  • Agentic AI
  • Quant
  • Governance

LLM-Native Engineering Workflow

Claude + Codex on a shared Obsidian prompt / rules layer across the RGC team — ~3× throughput, with review gates and compliance canaries keeping agent-generated code auditable.

  • LLM
  • Developer Productivity

Event-Driven Nano-Service Framework

Kafka on EKS with idempotent-keyed DynamoDB I/O — teams compose service DAGs without tight coupling or big-bang releases. Adopted by the ML team as Capital Connect's integration fabric at J.P. Morgan.

  • Kafka
  • EKS
  • Distributed Systems

Signal Platform Rewrite

Led a 6-developer rewrite from F# monolith to containerised Python on Kubernetes — 120 macro signals live, runtime days → hours, deploy weeks → one day, legacy stack retired.

  • Kubernetes
  • Python
  • Migration

SRE Turnaround

Drove firm-wide MTTR from ~8 hours to under one as SRE Manager at Winton; shifted ops from reactive firefighting to proactive monitoring, before “you build it, you run it” was standard.

  • SRE
  • Reliability