AI · Agentic delivery
A self-healing agent delivery system
Meta-agents, adversarial cross-engine review and proof-of-work gates that let autonomous coding agents ship real software without a human babysitting every step.
London · Product & Technology · Building AI
Luke Czak
Czak — “chahk” /tʃak/
Senior product & technology leader who never stopped writing code — now building the agentic-AI systems that build software.
Thirteen years scaling regulated financial products at HSBC, Shieldpay, Zopa and Irys. Today: autonomous multi-agent delivery, hands-on.
Available for select senior product & AI-engineering roles — permanent or contract.
About
A product leader with an engineer’s hands — and an engineer’s standards.
For over thirteen years I’ve worked at the seam where product meets engineering: building and scaling regulated financial products, B2B SaaS platforms, ML-powered features and internal developer tooling — from Series A startups to a global bank. I’ve run product and technical-product organisations of 50–80 people as a manager of managers, and I’ve been the sole product-and-technology decision-maker taking a startup from zero to launch.
I started as a software engineer and never put the keyboard down. I still read production code, lead technical design reviews, and care about the unglamorous things — security, accessibility, data integrity — that decide whether a regulated product is trustworthy. That mix is rarer than it should be: someone who can hold a board conversation about strategy in the morning and review a diff for an IDOR hole in the afternoon.
Most recently I founded my own company to go deep on AI. I build with autonomous coding agents every day — and I’ve shipped the hard part: a suite of provider-agnostic developer tools end-to-end as sole architect, plus several full-stack native ventures (iOS / watchOS, Android, FastAPI and Next.js). The real work isn’t prompting a model — it’s the engineering discipline that makes a swarm of agents reliable, honest and secure enough to ship.
A few things I’ve helped build
£5–8B
processed each year on the regulated payments platform whose
risk & rules engine I owned — Shieldpay.
850k+
customers on the FCA-regulated neobank where I
led mobile, web & API product — Zopa.
20+
markets composing their banking apps on the
HSBC mobile platform I owned.
50–80
people in the product & technical-product org I ran as a
manager of managers at HSBC.
0→20k
users I took a creative-tech product to as
founding CPTO, 0→1 in 9 months — Irys.
+25%
engineering efficiency from the
modular architecture I built the case for at HSBC.
Experience
A thirteen-year line — from first commit to a one-person AI studio.
-
2026 — Now
London
AppayaFounder & AI Engineering Lead
My own studio for AI & agentic software engineering.
Founded to go deep on agentic software development — building provider-agnostic developer tooling and the methodology that makes autonomous multi-agent delivery reliable, while staying hands-on as principal engineer.
- Orchestrate AI coding agents (Claude Code, Codex, Cursor, Kimi) to plan, build and review software under human direction.
- Engineered adversarial cross-engine review + proof-of-work gates that catch production-class defects — SSRF, IDOR, auth gaps, races — single-engine review misses.
- Shipped a suite of developer tools end-to-end as sole architect, plus full-stack native iOS/watchOS & Android ventures.
-
2024 — 2026
London
IrysChief Product & Technology Officer
A consumer photo & media platform — where people upload, share and discover images and video.
First product-and-technology hire — sole decision-maker on architecture, roadmap and engineering, with no CPO, CTO or Head of Engineering above or beside.
- Took the product 0→1 to public launch in 9 months, strategy through execution.
- Grew from zero to ~20,000+ users.
- Shipped an ML content-moderation pipeline (NSFW + text scanning) inside the upload flow at scale.
-
2023 — 2024
London
ShieldpayHead of Technical Product
Regulated B2B payments & escrow for legal and M&A deals.
A platform clearing £5–8B a year, serving 40+ of the UK’s top-100 law firms across escrow and disbursement.
- Owned the rules engine; introduced ML risk scoring for KYC/AML — manual review to automated, auditable workflows.
- Delivered a 200% integration-throughput gain on cross-border payments; cut onboarding queries 50% via self-serve.
- Rolled out OKRs and Agile org-wide — 80% sprint accuracy within four sprints.
-
2022 — 2023
London
Head of Engagement · Product Lead, Mobile & Web
An FCA-regulated UK neobank — lending, savings & money tools.
850k+ customers, a £2.1B loan book and £2.9B in deposits. Led 3 product managers across app, website and APIs.
- Owned the open-banking data product — spend categorisation, credit-score explanations, personalised nudges.
- One feature reached 5% weekly returning MAUs in its first month; lifted financial-product cross-sell +1.5%.
- Ran an A/B programme across the 850k-user base to validate decisions with data.
-
2021 — 2022
London
SmarketsHead of Product
A London betting & prediction-market exchange.
Head of Product for the exchange — owning the mobile and web betting experience, the markets and pricing surfaces, and the experimentation roadmap, between HSBC and Zopa.
-
2016 — 2021
London · Global
Global Head of Technical Product Management
The global digital arm of one of the world’s largest banks.
A digital-banking estate serving millions across 20+ countries. A manager-of-managers role — ten technical product managers within an org of 50–80 on a multi-million-pound budget.
- Owned the internal mobile platform — a shared component library letting 20+ markets compose banking apps without rebuilding.
- Orchestrated the new HSBC UK app launch, then US, Canada and beyond — aligning Risk, Legal and Architecture.
- Improved engineering efficiency 25% via a modular, reusable architecture.
-
2015 — 2016
London
GoustoLead Mobile Developer, also acting as Mobile Product Manager
A UK recipe-box / meal-kit subscription service.
Led mobile engineering for the meal-kit app and stepped into acting Mobile Product Manager — the hands-on build-it-and-own-it pattern that has run through everything since.
-
2012 — 2015
Wrocław · UK
ObjectivitySoftware Engineer
A bespoke software-engineering consultancy (UK & Poland).
Where it started — the hands-on engineering foundation under everything since. BEng in Computer Science (final grade 4.5/5), Wrocław University of Technology.
Now · Going deep on AI
I build with autonomous agents — and engineer the discipline that makes them trustworthy.
Software is increasingly written by AI. The hard part isn’t getting a model to generate code — it’s making a swarm of autonomous agents deliver reliably, honestly and securely. That’s the problem I work on, hands-on, every day.
Adversarial, cross-engine review
One model builds; a different model attacks the diff. Paired with screenshot and proof-of-work gates and hard anti-fabrication enforcement, it catches real production-class defects — SSRF, multi-tenant / IDOR isolation holes, forged receipts, auth gaps, races — that single-engine review waves through.
Meta-agents that supervise agents
Agents that watch other agents — observing live sessions, diagnosing stalls and verifying proof-of-work before allowing “done” — turning unreliable autonomous workers into a self-healing, walk-away delivery system.
Provider-agnostic by design
No single LLM vendor hardcoded. A multi-engine router and credential/quota broker routes work to whichever model is best, cheapest or available — with local-first, on-device options where data shouldn’t leave the machine.
In practice this has let me ship a string of developer tools and several full-stack native-app ventures solo — production code reviewed daily, not vibes — while the agents do the typing.
Claude CodeCodexCursorKimi Multi-engine orchestrationRAG & LLM patterns On-device LLMsPython · FastAPINext.js · React Swift · Kotlin
Selected work
A few things I’ve built and shipped.
AI · Developer tooling
A provider-agnostic AI toolchain
A growing suite, built end-to-end: a file-based task/decision tracker, a schema / design-token / i18n codegen + integrity gate, a multi-engine CLI router with a quota broker, an agent-runtime skills toolkit, and more.
Fintech · Platform
HSBC internal mobile platform
A shared component library that let engineering teams across 20+ markets compose market-specific banking apps without rebuilding — core developer tooling at global scale.
Fintech · Risk
Shieldpay risk & rules engine
KYC/AML decisioning moved from manual review to automated, auditable risk workflows — inside a platform clearing £5–8B in payments a year.
Fintech · Open banking
Zopa open-banking product
Spend categorisation, credit-score explanations and personalised financial nudges driven by connected-account data — validated by A/B testing across 850k+ customers.
0→1 · ML
Irys, zero to ~20k users
Sole CPTO taking a consumer photo & media platform from nothing to public launch in 9 months, including an ML content-moderation pipeline in the upload flow.
Outside the work
Music production and DJing, drone photography, and documented training.
The same appetite for craft and reps that I bring to building products — applied to a pair of decks, a camera drone and the barbell. Polish by birth; I work in English and Polish.
Contact
Available for select senior product & AI-engineering roles — permanent or contract.
London-based, hybrid or fully remote. If you’re building something at the edge of product, fintech and applied AI, I’d like to hear about it.