Mike Woodcock · Profile2026 · for startups & growth-stage teams

AI that survives contact with production.

Independent AI & engineering consultant. Agentic automation, LLM products and deep learning — designed, shipped and handed over so your own team can run them.

Based

Berlin · CET (UTC+1)

Works with

US & EU teams, remote

Contact

me@mikew.io

About me

The long version ↗

Mike Woodcock · independent AI & engineering consultant01

The short version02 — Framing

I take AI from “we should do something with this” to a number that moved.

What I am not: a slide-deck strategist, or an agency that ships a demo and leaves.

What I am: a technical leader who has run AI and engineering departments, and a senior engineer who still writes the code, runs the evals and carries the pager until your team can.

Profile02

Track record03 — Four numbers

What “shipped to production” has meant so far

€700K+

a year in overstock fees gone — demand planning automated, forecast error down a third.

60%

of support tickets resolved by agents, 2k+ a month, six-nines uptime.

20 → 200 fps

camera perception, 100+ live streams.

€100M+

of yearly revenue on infrastructure my teams owned, across 30+ marketplaces.

All from my own engagements. Where a client's numbers are confidential, I say so rather than round up.

Track record03

ServicesThree services04 — Three services

Three ways to work with me

Hands-on by default. Each of the three stands alone, and most engagements move from one to the next: a project that proves the value, then a fractional role that makes it stick.

1 · Agentic automation

Agents that read your data, use your tools and finish the job — in support, operations, finance, sales or engineering — with a person in the loop where it matters.

Proven at unicorn scale

2 · Deep learning & computer vision

Models that see — and that edit: detection, tracking, depth and video understanding, plus generative image and video for VFX, compositing and restyling. Trained on your data, made fast enough for the hardware you actually have.

Shipped inside a Ford plant

3 · AI lead

AI strategy and senior judgement, one or two days a week, until you can hire it full time: where to place bets, what to build versus buy, how the organisation should change, and how to tell the story to the board.

For: AI strategy and roadmap, operating model and governance, build-vs-buy and vendor reviews, architecture reviews, hiring and mentoring, the shared platform every team builds on.

Former Head of AI & Engineering

Services04

AI transformationStructure05 — Two altitudes

Two altitudes, one direction

The company - Leadership

AI transformation. The strategic and leadership vision: where the company is going with AI, which bets to make, who owns what, and how success is measured.

The engineering: The Execution

Agentic software factory. The implementation: the low-level details, the plan and the execution — how the work actually gets built, checked and shipped.

The transformation creates the demand; the factory creates the supply. Run only the first and engineering is the bottleneck. Run only the second and you ship faster towards nothing in particular.

AI transformation · structure05

AI transformationConcept one06 — The company

Adoption that shows up in the P&L, not in a slide

From scattered experiments to a portfolio of automations with owners, numbers and one platform underneath. Not a tool roll-out — an operating-model change with a technical core.

Done before: SellerX, ten markets, three continents. 60% of support tickets and most of demand planning automated on a stack every team shared.

1

Portfolio

Every process, ranked by value × feasibility. Two or three lighthouses with a baseline number.

2

Foundations

One governed stack, so five teams don't build five of everything.

3

Lighthouses

First two use cases live within 90 days, each with an owner and a number that moved.

4

Operating model

Decision rights, training, measurement — so it keeps spreading after I leave.

AI transformation · concept one06

AI transformationConcept two07 — The agentic constellation

The agentic constellation

A company where agents do the work

One inbox, six manager agents, a few hundred working agents a week. They take tasks, do them, check each other's work and report back. Anything above a limit, a spend, a send, a deploy, stops and waits for a person.

The next slide is a real week of my own setup, drawn from its event log and replayed at ten hours a second: every dot an agent, every line a message, a question for me or a decision.

Background: the same week, running behind this text. Blue lines are my messages going out, white lines results coming back, red an agent waiting for me, green a decision, amber a launch or a deploy.

AI transformation · the constellation07

AI transformationConcept two08 — The agentic constellation

AI transformation · the constellation · a real week, replayed08

AI transformationConcept three09 — The agentic software factory

A coding agent speeds up one developer. A factory speeds up delivery.

Coding agents make implementation faster. Delivery still waits on review, decisions, hand-offs and release. More output into the same system just moves the bottleneck.

69%

of developers say agents made them more productive

17%

say agents helped their team work better together — Stack Overflow Developer Survey 2025

A software factory is a delivery system: intent in, bounded work for agents, checks before anything ships.

Intent, rules and gates live in the repo, not in someone's chat.
Agents run whole lifecycle steps on a trigger, not just autocomplete.
Checks decide what proceeds; people decide what exists.

AI transformation · the factory09

Part two10 — Selected work

Selected work

Things that are running right now

Three slides: the unicorn, the car factory, and the 2026 engagements. Numbers are the ones I'm allowed to share.

Background: my own recording at IAA Mobility, Munich, September 2021. The Kopernikus automated-valet-parking system driving a BMW iX through the demo garage with nobody inside, on cameras alone: no lidar, no driver. I owned that camera perception end to end, from the models to the calibration to the on-vehicle deployment, and was on site when this was filmed. More on the Kopernikus slide.

Selected work10

Selected workSellerX · Berlin · 2024–202611 — E-commerce

Automated the planning and support work behind €100M+ of yearly revenue

Role: Senior AI Architect → Head of AI & Automation → Head of AI & Engineering. 100k+ products, 30+ marketplaces, 100+ brands.

Built: the forecasting platform for every article, the agents that answer customers, and the internal hub half the company works in.

70–80%

of demand planning automated; four planners before

−30%

forecast error, halved on the long tail

3.5×

more new listings for targeted brands

6 nines

uptime on the ticket agents, 2k+ a month

Selected work · SellerX11

Selected workKopernikus Automotive (acquired by Ford) · Berlin · 2019–202412 — Autonomous driving

Camera-only autonomous driving inside a live Ford factory

ML Engineer → ML Lead → Senior ML Engineer. Camera perception owned end to end, from design domain to the code on the vehicle. No lidar, no driver.

20 fps

200 fps

100+ camera streams, live · IAA Mobility 2021, my recording

Millimetre accuracy from calibrated cameras and geometry alone.
10× faster perception on the hardware already in the plant.
Commissioned on site, validated against production data, safe roll-out and roll-back.

BMW, CARIAD, Ford, Jaguar, Land Rover and Mercedes-Benz on the IAA 2021 partner wall — IAA Munich 2021 · the partner wall

Selected work · Kopernikus / Ford12

Selected workIndependent13 — Current clients

Some of the clients since going independent

San FranciscoUTC−7

BogotáUTC−5

Berlin · meUTC+1

Frankfurt · LorschUTC+1

Television.AIFrankfurt

AI video editing for newsrooms: ingestion, transcription, knowledge extraction, and models that learn each channel's editorial voice.

Status: in use at several of Germany's leading public broadcasters.

Fractional AI lead

AudioscrapeSan Francisco

The ingestion pipeline behind a searchable index of 1M+ hours of podcasts: resilient, fast and cheap to run.

Consulting engineer

Traducciones 24/7Bogotá

An AI-native translation company, designed and built from the ground up: a constellation of agents runs intake, document extraction, pricing and quoting, with people only where a sworn translation needs one. Installed, handed over, and quoting in seconds instead of hours.

Agentic architecture

Seneca EvolutionWholegrain Finance · Lorsch

Strategy for a financial-controlling product: one clean data model, retrieval on top, new customers onboarded with almost no manual work.

Technical strategy

Plus advisory work with founders on AI roadmaps, build-vs-buy, architecture reviews and hiring. References for any of the above on request.

Selected work · 202613

Part three14 — About me

About me

Ten years of turning research into things that run.

Mathematician by training, engineer by temperament, economist enough to care about the P&L.

UK (Edinburgh) · Germany (Berlin) · Latin America (Colombia). Studied mathematics, systems engineering and economics in Bogotá; MSc in Artificial Intelligence in Edinburgh; four and a half years of autonomous driving in Berlin, then a unicorn's AI and engineering departments. Independent since 2026.

University of EdinburghUniversidad Nacional de ColombiaOxford ML Summer SchoolKopernikus Automotive → FordSellerXFormula Student AI · 1st place

The full story, credentials and stack →

About me14

About meThe path15 — 2012 → 2026

From engineering
to expert AI consultant.

2012

Systems engineering, then mathematics, with merit

Universidad Nacional de Colombia · Bogotá · six scholarships for academic performance

Started in systems engineering, added mathematics for the rigour, and won the scholarship for academic performance every year I was there: six times, from the engineering, science and economics faculties. Engineering taught me to build, mathematics to prove.

2017

Economist, with merit

Universidad Nacional de Colombia · BA Economics · top 5 of the class

The degree I finished. Economics asks the only question left once something is built and proven: did the number move? I still work in that order.

Top 5 of the 2017 class

2018

University of Edinburgh with Merits

MSc Artificial Intelligence, with merit · perception for the Formula Student driverless car

The year the theory got wheels. I owned what the driverless race car could see: cameras in, track out, running live on the vehicle. The car won.

1st place, UK autonomous-driving competition 2019

2019

ML Engineer → ML Lead, camera perception

Kopernikus Automotive · Berlin · autonomous driving

Cars that drive themselves on cameras alone: no lidar, no driver. I owned perception end to end, from design domain to the code on the vehicle, and took it from 20 to 200 frames a second on the hardware already in the plant.

10× faster perception, 100+ live camera streams

2021

Senior ML Engineer, on site at IAA Munich

Kopernikus Automotive · the partner wall: Mercedes-Benz, VW, BMW, Jaguar, Ford

A BMW iX drove itself through the demo garage with nobody inside, in front of the whole industry, on my camera stack. Oxford ML Summer School the same summer. Ford went on to acquire the company.

Live in a Ford plant, camera-only

2024

Senior AI Architect

SellerX · Berlin unicorn · 100k+ products, 30+ marketplaces, 100+ brands

Hired to put AI under a company that runs on numbers. I designed the forecasting platform for every article and the first agents that answer customers, and shipped them myself rather than handing over a diagram.

−30 % forecast error, halved on the long tail

2025

Head of AI & Automation → Head of AI & Engineering

SellerX · Berlin · two steps up in two years

Ran the AI and engineering departments: the platform every team builds on, the agents answering 2k+ tickets a month at six-nines uptime, and the hub half the company works in. Planning and support behind €100M+ of yearly revenue.

70–80 % of planning automated · 60 % of tickets resolved by agents

2026

Independent AI & engineering strategic consultant

Berlin · US and EU startups, remote

Now I do it for a few companies at a time: fractional AI lead for a newsroom tool, architect for a translation business, consulting engineer for a podcast search engine, adviser to a private-markets platform. Designed, shipped and handed over, so your team runs it without me.

Clients on three continents

About me · the path15

Next step17 — Contact

Tell me what you're building.

A short email is plenty.