Twenty years in industrial digitalization. Six frames.

I spent my twenties on a question that sounded academic at the time: how should people handle 3D content without a mouse? Touch, sketches, gestures. Five published papers, a doctorate at INRIA, a short postdoc in Tokyo. It stopped sounding academic the moment factories wanted the answer.

The route there was deliberately mixed: Weimar, a design school that happened to teach computer science, then a vision lab in San Sebastián, a materials prototype at Daimler in Ulm, three years at INRIA, two months in Tokyo. Four countries and one question — plus the habit of testing an idea on someone who has no reason to be patient with it.

• 5published papers
• 1best paper award
• 4countries, one question

Work & research archive ↗

I studied computer science at the Bauhaus-Universität Weimar, a school known for design and engineering — which is exactly why the interface questions found me there. Media, computation and form sat in the same building, and the interesting problems sat between them.

• 2001–2006Diploma in Computer Science, Bauhaus-Universität Weimar — equivalent to an MSc
• 2003–2004Research intern, Vicomtech-IK4, San Sebastián — augmented environments
• 2005Research intern, Daimler AG, Ulm — scalable material model

My doctorate at INRIA asked how people could handle 3D content without a mouse. The thesis proposed interaction techniques that stay simple while the data gets complicated: single and multi-touch input driving full 3D manipulation, strokes and sketches instead of menus, and illustrative rendering that highlights what you are working on without throwing the context away. Every technique went in front of real users and was measured. A technique that only works for its author is not a result.

• 2006–2009PhD in Computer Science, INRIA / Université Bordeaux I
• ThesisBeyond Desktop: Designing Novel User Interfaces to Enhance 3D Exploration

Read the thesis ↗

A CNRS grant took me to the Hirose-Tanikawa laboratory at the University of Tokyo at the end of 2009 — a short, dense stay in one of the strongest virtual-reality groups anywhere. Back in France I spent a year as ATER at the Université de Bordeaux, teaching computer science while finishing the research. Teaching is where I learned to explain a technical idea to someone who has no reason to care yet, which turned out to be the job description for the next fifteen years.

• 2009Post-doctoral researcher, University of Tokyo — Hirose-Tanikawa lab, on a CNRS grant
• 2009–2010ATER, Université de Bordeaux — teaching and research

Five papers, and a Best Paper Award at IEEE 3DUI in 2008 for Navidget.

• 3DUI 2008Navidget for Easy 3D Camera Positioning from 2D Inputs — Best Paper Award
• Multi-touchMulti-touch RST in 2D and 3D Spaces: Studying the Impact of Directness on User Performance
• HologramsInteracting with Augmented Holograms
• OpticsConsistent Illumination within Optical See-Through Augmented Environments
• Mobile VRSketch-based Route Planning with Mobile Devices in Immersive Virtual Environments

At Diota I led the R&D team that turned AR tracking research into a product carmakers actually use. We developed the core technology with CEA’s vision lab, then hardened it for real factories — bad lighting, moving parts, operators with no patience for toys. The tech survived. That’s the part I’m proud of.

I arrived as a project manager, which in a small company means owning the outcome rather than the schedule. Three years of putting AR into assembly, quality control, maintenance and training taught me where it genuinely earns its place, and that was the brief I took into the R&D directorship in 2013.

• 7years, two roles
• 2industrial solutions launched
• 1tracking stack industrialized

I joined Diota in Paris as a project manager, which in a small company means you own the outcome as well as the schedule. We put augmented reality into assembly, quality control, maintenance and training for French and German manufacturers. The best part happened on the shop floor: finding the step in a process where AR genuinely earned its place, then proving it with the people who would use it every day.

• 2010–2013Project Manager, Diota, Paris — automotive and aerospace clients in France and Germany

As Director of Research and Development I ran the cross-functional teams building the AR and computer-vision product. We industrialized the tracking technology for automotive, aeronautics and consumer goods, and launched two AI-driven industrial solutions with top-tier OEM clients. The standard we held to: a demo should work just as well in someone else’s hands as in ours.

• 2013–2017Director of Research & Development, Diota S.A.S.

The core was constrained SLAM tracking, developed with CEA’s Vision Engineering Lab and referenced against the CAD model, so the system knows what it is looking at rather than inferring it from features alone. Then came the craft — making it hold up under factory lighting, around moving parts and reflective surfaces, in the hands of operators who need tools that simply work. That hardening is what carried a strong research result into technology still in use today.

• On the lineAssembly guidance, quality control, maintenance, training
• SectorsAutomotive, aeronautics, consumer goods

Between 2015 and 2017 I took an Executive MBA in Management and IT at EPITECH Paris, in the evenings, while running R&D. I wanted the vocabulary of the business case to sit alongside the engineering: how an investment gets justified, what a finance director needs to see, how to make the case for a product as well as build it. It turned out to be exactly the preparation for the job I took next.

• 2015–2017Executive MBA, Management & IT, EPITECH Paris

In 2017 I took over Diota’s German company. New market, small team, my name on the paperwork. Over five years we won major automotive and aerospace accounts across DACH and grew revenue year after year. Selling AR to German industry taught me two things worth having: patience, and exactly what a business case needs to look like before anyone signs.

Running a company is a different job from running a team: hiring, the P&L, and a signature that carries legal weight. What I kept from it is the habit of asking what a number has to look like before anyone will act on it — a question that matters as much in engineering as it does in sales.

• 300%revenue growth
• 45%CAGR, DACH & Eastern Europe
• 5years as Managing Director

I took over Diota GmbH as Managing Director in 2017, based in Stuttgart — profit and loss, hiring, and the responsibility that comes with the title. The French parent had the product; Germany had the customers who would prove it in production. No playbook existed for selling industrial AR into this market, so we wrote one and sharpened it every quarter.

• 2017–2022Managing Director, Diota GmbH, Stuttgart

Selling AR to German industry is a long game. The buyers are engineers, the approval chain is long, and nobody signs because a demo looked impressive. We segmented by process rather than by industry: find the plant with the specific problem the technology genuinely solves, prove it there, and let that reference travel. That approach covered DACH and Eastern Europe, across automotive, aerospace, machinery and research institutes.

• TerritoryDACH and Eastern Europe, built from a standing start
• SectorsAutomotive, aerospace, machinery, research institutes

Revenue grew 300% over four years and compounded at about 45% a year across the territory, on value-based selling and careful segmentation rather than discounting. The measure I care about is not the multiple but the shape: growth that came from repeatable use cases, so the business could be handed on intact.

• Growth300% revenue growth in four years
• Rate+45% CAGR across DACH and Eastern Europe

Volkswagen, Porsche, Airbus Helicopters and Rolls-Royce became accounts in those years, alongside machinery manufacturers and research institutes. Each one started the same way — a single process, in a single plant, with a number at the end of it — and grew from there.

• AutomotiveVolkswagen · Porsche
• AerospaceAirbus Helicopters · Rolls-Royce

Today I’m part of the DELMIA Augmented Experience consulting team at Dassault Systèmes. We help automotive and aerospace clients connect manufacturing planning with the shop floor — the digital twin on one side, the people doing the work on the other. My job is making sure the two actually meet.

I joined in 2022 as a solutions consultant and have led the consulting team since June 2024. The problem rhymes across every client: planning is accurate the day it is written, and production never stops moving. Closing that distance is less a software question than one of who is trusted to change the plan, and how quickly.

• 7consultants in the team
• >50%of revenue 2023
• −18%client production time

DELMIA Augmented Experience ↗

I joined Dassault Systèmes in 2022 as a solutions consultant for the DELMIA Augmented Experience portfolio. Consulting is where the useful conversation happens early: what the software does, and what it takes in a customer’s own process to get the most from it. Having that conversation up front accelerates sales cycles.

• 2022–2024Solutions Consultant — industrial process consulting

Since June 2024 I have led the consulting team for DELMIA Augmented Experience — seven consultants delivering AR and AI-based solutions to industrial clients. The measure of a good engagement is what the customer keeps: a way of working their own people can run with confidence once we hand it over.

• since 2024Consulting Team Manager, DELMIA Augmented Experience

The work gets measured on the floor rather than in the demo room. Targeted implementations brought client production times down by up to 18%, and the workflows we optimized gained up to 25% in efficiency. In 2023 the accounts I supported carried more than half of revenue. A digital twin only pays for itself when the plan on the screen matches the work being done.

• ProductionClient production times down by up to 18%
• WorkflowsEfficiency gains up to 25%
• Revenue>50% of revenue in 2023

BAE Systems, MAN, Daimler Truck, Spirit AeroSystems, Airbus and Safran, among many others — aerospace, automotive and truck manufacturers connecting manufacturing planning with what actually happens on the line. The problems rhyme across all of them: the plan is accurate the day it is written, and the floor keeps moving.

• AerospaceBAE Systems · Spirit AeroSystems · Airbus · Safran
• AutomotiveMAN · Daimler Truck

Alongside the day job I design and ship software end to end. StockScore rates companies across 100 points and shows every calculation behind the number. Wattlas explains European electricity from open data and rebuilds itself on a schedule. Both are live, and both were built the same way: I set the plan and approve the release, Claude does the building.

Neither one is a demo. Both run on a schedule, carry their own tests, and keep their credentials out of the repository — the same bar I would set for anything going near a plant. Building them is how I stay honest about what AI-assisted engineering actually delivers, and where it still needs a human to say no.

• 2products, both live
• 2human gates: plan & release
• 0secrets in either pipeline

StockScore rates a company out of 100 and shows every calculation behind the number: 45 points of fundamentals, 30 of valuation, 25 of moat. The moat assessment runs the business summary through Hamilton’s 7 Powers with AI assistance; the fundamentals are nine checks on growth, margins, balance sheet and how well cash is reinvested. Around the score sit deliberately score-free layers — a sentiment radar, an IPO radar, hand-curated event briefs — because not everything worth knowing should be compressed into a single number.

• StackNext.js · FastAPI · Cloud SQL · Gemini on Cloud Run

Open StockScore ↗Read the case study

Wattlas explains European electricity from open data: 35 views organized around six editorial questions instead of technical categories. Why power is cheap at midday and expensive at seven. What is actually on the grid. Where geography sets the price. There is no backend — a Python pipeline reads ENTSO-E, NESO, MaStR, SMARD, Eurostat and others at 05:17 UTC, writes JSON files, commits only what changed, and the static site redeploys.

• Tests156 offline tests — no network needed to prove the pipeline works

Open Wattlas ↗Read the case study

Both products ship through the same operating model. I write the tickets and set the plan; Claude builds — code, tests and mockups — and I approve the release. Work moves in small vertical slices, one issue per git worktree so parallel sessions stay clean, and any significant approach gets stress-tested against a second model — twice — before a line of code exists.

How it ships, in full