Skip to content
Currently available — for the right work·France903+ Day French Streak·2026 Q2 calendar — open now
Lead Product EngineerLondon · full-time, in-officeSourced via Dex

You simulate the opinions that matter. I have spent nine years shipping the product layer that decides whether anyone can act on them.

This is your first lead-level product engineering hire, and the JD is explicit about what that means: architectural calls on data models and API contracts that hold for years, and the latency, the empty states and the small interactions that decide whether a feature feels right — both axes, at once, on a product Fortune 500 clients make high-stakes decisions on. That is the job I have been doing since 2017: co-founder to CIO at , 0→1 across five verticals, and for the last year sole engineer on two systems of my own where the schema, the API and the pixel are all mine to get right.

Fauzul Kabir Chowdhury
Lead Product Engineer, applying

Co-founder → CIO at a venture studio for nine years; now sole engineer on two production systems. Ships the vertical slice — browser to Postgres — and sets the bar the rest of the team builds to.

London-bound9 yrs shippingTypeScript nativePython secondFounderClaude-led build
9 yrs
Shipping to real users
Co-founder → CIO at ELO, 2017–2026. 0→1 across music-tech, gaming, hospitality, health-tech and professional networks.
10
Client geographies
Dhaka · Zurich · Möhlin · Oslo · Sandnes · London · Dubai · California · New York · Quebec — delivery, not travel.
15,000+
Users on one checkout
Jamahook, Zürich — six years on one codebase through Stripe, GDPR and EU VAT, to a Loopcloud acquisition in 2026.
1,863
Tests passing
Wiregent, built solo and live at wiregent.com. In one eight-day sprint the test files grew faster than the source they cover.
22
Engineers led
Grew ELO’s team to 22, then ran it remote-first. Two alumni now lead teams in Norway.
HOW I DIFFERENTIATE

Where a lead product engineer usually sits

Your JD names the tension outright: architecture that shapes the product for years, and empty states that decide whether it feels right. Most engineers are honestly one or the other.

PURE PRODUCT

Interaction-led engineers

  • Sweat latency, empty states, the small interactions
  • Fast on the surface a client actually touches
  • Data models inherited, not designed
  • Contracts drift once other systems depend on them
WHERE I SIT

Both axes, same week

  • UX Engineer first, principal engineer after — in that order
  • Schema, API contract and pixel owned together
  • Codegen so the contract cannot drift by hand
  • Nine years of the reliability half, on client-facing work
PURE INFRASTRUCTURE

Systems-led engineers

  • Service boundaries, query plans, queueing
  • Schemas designed to last
  • Surface handed to somebody else
  • The feature ships correct and feels wrong
The simulation

6 requirements, run one at a time

Your JD runs six bullets. Each one is below as a simulation run: the requirement as the question, the verdict as the badge, and real responses underneath — evidence on one side, and where a ramp is honest, the ramp.

Requirement 1 of 6
Track record

At least 5 years shipping product to real users, owning features from idea through to production, and you've been the engineer others turn to when the problem is hard

Strong fit
Evidence
Shipped workTrack record
Response
Strong fit
Comment

Nine years at , co-founder through CIO, shipping 0→1 platforms across five verticals for clients in ten geographies. is the clearest single artefact: an API marketplace with authentication (Auth0), RBAC and multi-instance tenancy, with Elobooks riding on it as the consumer product — one platform, two audiences, one schema. Since stepping out of the operating role in August 2026 I have been the only engineer on and Trestate, which is the hard-problem role with nobody else to turn to.

Why it holds
No caveat needed
Response
Answered with work already in production
Comment

Nothing on this line is a projection. Every system named above is running, and the case study behind each one is a click away in the chat.

5 of 6 answered with shipped, checkable work.
Lead strength

The vertical slice is the unit I actually work in

Trestate is the cleanest proof: I chose the Go monolith, defined the ~200-feature vector, wrote features.yaml as the single source of truth that codegens SQL, Go and TypeScript, built the Python + DuckDB + Dagster pipeline that normalises four national datasets into it, and then built the React 19 + deck.gl frontend that renders every scored hex interactively. Front-end interaction through API to database, one person, one system — which is the ownership your JD describes.

Introduction to Agent Skills · Anthropic · VerifyClaude Code 101 · Anthropic · Verify

The vertical slice is the unit I actually work in

Bring me in for the vertical slice. Give me a real feature with a front-end, an API contract and a schema change in it, a week, and access to the engineer who knows where the bodies are buried — and judge the pull request, not the CV. That is how I would evaluate this hire if I were you, and it is the format I am fastest in.

AI coding agents held to a shipping bar, with the certificates to match

Bring me in for the vertical slice. Give me a real feature with a front-end, an API contract and a schema change in it, a week, and access to the engineer who knows where the bodies are buried — and judge the pull request, not the CV. That is how I would evaluate this hire if I were you, and it is the format I am fastest in.

Strength 02

AI coding agents held to a shipping bar, with the certificates to match

Anthropic's Claude Code 101 and Introduction to Agent Skills, both verifiable, plus Google × Kaggle's 5-Day AI Agents Intensive. The practice behind them: 's team writes plain English, Claude opens the pull request, CI/CD ships it, and every change stays reviewable and revertible in Git. The bar is enforced by tests — 's 1,863-test suite exists so agent-written code can be merged without faith.

Strength 03

Reliability treated as a product property, because clients bet on it

Your JD says clients make high-stakes decisions on this product, so reliability isn't negotiable. My version of that: reports a pre-revenue company as unproven rather than declining it for lacking a metric it does not have, and its ledger throws on an unbalanced leg set rather than appending; Trestate is licensed to say a position resembles where a trade is established and is forbidden from predicting success. Both are engineering decisions about what a system is allowed to claim.

Reliability treated as a product property, because clients bet on it

Bring me in for the vertical slice. Give me a real feature with a front-end, an API contract and a schema change in it, a week, and access to the engineer who knows where the bodies are buried — and judge the pull request, not the CV. That is how I would evaluate this hire if I were you, and it is the format I am fastest in.

Standards set by review, on teams that grew

Bring me in for the vertical slice. Give me a real feature with a front-end, an API contract and a schema change in it, a week, and access to the engineer who knows where the bodies are buried — and judge the pull request, not the CV. That is how I would evaluate this hire if I were you, and it is the format I am fastest in.

Strength 04

Standards set by review, on teams that grew

went from a founding team to 22 engineers, then remote-first from 2020 with a 75% reduction in cycle time via GitHub Actions CI/CD and automated multi-cloud deployments. Compliance was baked in across five portfolio platforms — HIPAA, WCAG 2, GDPR — rather than retrofitted. Two alumni now lead teams in Norway, one is pursuing further education in Canada on my recommendation, and another is exploring their next chapter in Germany; one early-career direct report is now at Amazon in Sweden.

BEYOND THE CHECKLIST

The four you said you'd also value

The JD lists four things it would value and marks all of them optional. They are the reason to hire a lead rather than a senior, and I have all four.

01

User-facing production AI agents

runs a multi-agent n8n pipeline in daily production — ReAct-style reason-act loops scripting dual-host dialogue, with grounding evals gating every script against the scraped source before it reaches synthesis. fans a parsed profile out to six country agents running in parallel as FastAPI microservices under a Google ADK agent-to-agent orchestrator, grounded on live search because immigration policy drifts monthly. Both are agents real people use, not demos.

NewScriber · VisaPros
02

Managing infrastructure

is deployed and operated solo for about six dollars a month: Cloudflare Workers and R2 at the front door, a single Fly.io machine holding the event stream, Neon Postgres prepped behind a parity suite. Before that, nine years of cost-aware AWS and GCP — EC2 fleets justified with instance comparisons, container images in Lambda early enough to be an odd choice at the time, CloudFormation SAML provisioning in production.

Cloudflare · Fly.io · AWS · GCP
03

Seed to millions in ARR

went from a founding team I structured in 2017 to 22 engineers and a client list across ten geographies. was a direct Zürich client for over six years — React and NestJS, Stripe checkout for 15,000+ active users, GDPR and VAT — and was acquired by Loopcloud in 2026, with continuing as the technical team. I have been in the room for the whole arc, including the parts where it is unglamorous.

ELO · Jamahook
04

Mentoring engineers

I coached engineers into senior roles abroad: two alumni now lead teams in Norway, one is pursuing further education in Canada on my recommendation, and another is exploring the next phase of their career in Germany, with one early-career direct report now at Amazon in Sweden and alumni leading at two local Bangladeshi tech firms. The mechanism was never a mentoring programme — it was code review as the place standards are set, which is exactly what your JD says this role becomes as the team grows.

22 engineers · 4 countries
WHAT COMES BACK

The honest read

Your quality check returns a score, a band, coverage, findings and caveats — and it says out loud what it could not measure. Same format, applied to me.

Fit report

Self-assessed
5/6
Strong fit · one honest ramp
requirements answered with shipped work

Coverage5 of 6 requirements answered with shipped, checkable work

AI coding agents
First-party Anthropic certifications plus a client CMS running on plain English → pull request → CI/CD.
Whole-stack reasoning
Latency decided before schema on Trestate; WebGPU inference in the browser on fauzul.com.
Postgres depth
Hash-chained ledger, PostGIS at 72,863 hexes, HIPAA pipeline, multi-tenant RBAC.
Architecture that survived growth
Anygum's boundary held when ELO's own product had to consume it as a client.
Python depth
Real and second, not primary. Your engine is Python; I read and extend it, and close the rest on the job.

Caveat: This report measures fit against the six bullets you published. It does not measure what I do not yet know about your simulation engine — which is most of it, and which is the point of the first ninety days.

FIRST NINETY DAYS

How I'd construct the run

Your method page walks Persona Construction → Society Creation → Simulation Engine. Here is the same shape applied to the first quarter of this role.

1

Read the system as it is

  • Ship one full vertical slice in week one — front-end through API to schema — chosen for how much of the system it forces me to read
  • Sit with the forward-deployed engineers on a live client engagement before proposing anything
  • Write down where the Python engine and the product actually meet, because that boundary is the role
2

Set the bar in code review

  • Standards land as reviewed pull requests and a test suite, not as a document nobody reads
  • Make the latency budgets explicit and measurable per surface
  • Name the data-model decisions that are load-bearing for years and get them argued now, while the team is small
3

Compound it

  • Turn the repeating parts of the build into agent skills and harnesses the whole team can run
  • Zero-downtime migration path defined before the first schema change that needs one
  • Hand back a product surface a new engineer can ship into in their first week
What this page doesn’t do

The objections, before you have to raise them

The objections I would raise about my own application, answered before you have to ask them.

What this page argues
  • The vertical slice is the unit I actually work in
  • AI coding agents held to a shipping bar, with the certificates to match
  • Reliability treated as a product property, because clients bet on it
  • Standards set by review, on teams that grew
What it doesn’t claim
  • Python as a primary language, at the depth of your simulation engine
  • Simulation and network-science domain knowledge
  • Knowledge of your codebase — which is the point of the first ninety days
Your JD says expert TypeScript and productive Python. You've marked Python as your one ramp. Why should that not be disqualifying for a lead role on a Python simulation engine?

Because the JD splits them deliberately, and I think it is right to. The simulation engine is owned by your ML and research engineers; this role turns frontier capability into product, which is TypeScript-forward work with a Python boundary in the middle. I am expert on the first and productive on the second — Trestate's ingestion, ' FastAPI mesh and 's sandbox stage are all Python I wrote and maintain. If the honest requirement were 'expert Python', I would not be applying to lead.

You have been a founder and a CIO. Are you going to be happy as an individual contributor writing features?

It is the reason I am applying rather than raising. For the last year I have been sole engineer on and Trestate by choice — 247 commits and roughly 96,000 lines on one of them, alone, because that is the work I want to be doing. What I want back is a team around it. Lead product engineer is precisely the shape: still the person writing the hard slice, now also the person whose review sets the standard. I am not looking for a management title here and would say so in the interview.

You have never worked on simulations, network science, or synthetic populations. Isn't the domain gap real?

It is real and I will not talk it down. What I would push back on is the idea that it is the risk. The risk in a first lead product hire is somebody who can design a schema but ships a feature that feels wrong, or somebody with taste who leaves you a data model you regret in year two. That is the pair I can evidence. The domain I learn the way everyone does — by sitting on live client engagements with your forward-deployed engineers before proposing anything, which is the first thing on my ninety-day plan.

Five days a week in the office in London, and you are currently remote in Dhaka. How real is that?

As real as it gets — London has been top of my list for a while. The UK has been a delivery geography of mine for years, my sister lives in the city, and I would come in on a company-sponsored Skilled Worker route, which is the one I would prefer and the one I would ask you for. On the in-office question specifically, the remote years were a consequence of running a distributed firm, not a preference I fought for. A room and a whiteboard are what I have missed most since stepping out of ’s operating role.

Most of your recent work is solo. Your JD says the standard you set in system design and code review becomes the standard everyone builds to. Where is the evidence you can do that with other people?

The nine years before the solo year. I structured 's founding team, grew engineering to 22, moved it remote-first in 2020, and cut cycle time by 75% with CI/CD that people actually followed. The durable evidence is where those engineers went: two now lead teams in Norway, one is pursuing further education in Canada on my recommendation, another is exploring their next chapter in Germany, one is at Amazon in Sweden, and two lead at local Bangladeshi tech firms. None of that came from a standards document. It came from review being the place the bar was visible.

You use AI coding agents heavily. How do we know what ships is yours and is good?

Because the bar is mechanical, not a matter of trust. 's eight-day sprint added test files 17 → 47 against source files 101 → 222 — tests grew faster than the code they cover — and the suite is at 1,863 passing. On the client's plain-English request becomes a pull request that a human reviews before CI/CD ships it; nothing reaches production unreviewed. Agents make me faster at the parts that are typing. They do not get to decide what is correct.

Technical inventory

The stack, scoped to this role

The stack, scoped to what this role actually touches.

TypeScript & product surface
TypeScriptReactTanStack StartTanStack QueryWebGPUStorybookVite
APIs & services
Node.jsExpressNestJSPythonFastAPIGoRESTGraphQLWebSocket
Data
PostgreSQLPrismaPostGISDuckDBDagsterZod
AI engineering
Claude CodeAnthropic SDKAgent Skills & hooksGoogle ADKGeminin8nEvals
Infrastructure
Cloudflare WorkersFly.ioAWSGCP Cloud RunDockerGitHub ActionsVitest

Bring me in for the vertical slice.

Give me a real feature with a front-end, an API contract and a schema change in it, a week, and access to the engineer who knows where the bodies are buried — and judge the pull request, not the CV. That is how I would evaluate this hire if I were you, and it is the format I am fastest in.

London is where I want to build, and the UK is the country I have most enjoyed delivering into — years of client work there before moving was ever the plan, and my sister lives in the city. I would come in on a company-sponsored Skilled Worker route, which is the one I would prefer and the one I would ask you for. Five days a week in the office is a feature of this job for me, not a cost.

London · full-time, in-officeSponsorship welcome — Skilled Worker preferred
Fauzul Kabir Chowdhury · application fit
Application fit for Lead Product Engineer · Artificial Societies the posting