EUCLIDE/ About

I spent ten years on construction sites before I shipped software. Now I do the hard part of AI in engineering: getting teams who don't trust it to actually use it on real work. I've done that for 200+ engineers inside a global firm.

Know more about me

Who I am

I'm a civil engineer from Turkey, based in Paris. I spent ten years on real construction sites in seven countries, railways, bridges, tramways and water plants, before I moved into code. That field time is why engineers listen to me. I've stood where they stand, and I don't sell them tools I wouldn't use myself.

What I do, in plain terms

I help engineering teams use AI on real work. Day to day that is three things:

  • Find the right problem. I sit with the people doing the job and spot the tasks worth automating, the ones that eat hours and follow rules.
  • Build the tool. I ship it to production, and I build it so you can see what it did. Every claim carries its page.
  • Get people to use it. This is the part everyone skips. A tool nobody trusts is worth nothing. I train the team and stay until it sticks.

What I believe

Most AI in engineering fails as a demo that impressed someone once. The gap is never the model. It is trust and adoption. Close that, and the rest follows.

By the numbers

10
years on engineering & construction sites
7
countries on site
200+
CAD / BIM users onboarded
59+
countries with paying Styly users

The arc

2022 → now
AI & BIM adoption, then Applied AI · SYSTRA

Discovery with business teams, use-case prioritization, shipping production AI, and training the people who use it.

2021 → 2026
Co-founder, product & strategy · Styly

Built and ran an AI product end to end. Paying users in 59+ countries.

2021 → 2022
Digital deputy project manager · Assystem

Bridge between engineering, BIM and delivery on large infrastructure programs.

2014 → 2020
Site & construction lead · field

Railways, bridges, tramways, water plants across Algeria, Ethiopia and Saudi Arabia.

How I build

I build AI people can trust, because trust is what makes them use it. Every claim carries its page. The tools I ship can be read and checked, so a skeptical engineer can see exactly what the machine did. No black boxes.

PythonClaude / Claude CodeLangChain / LangGraphRAG & agentsAWS BedrockTerraformNext.jsOpenTelemetry

In the open

spec-triageOpen source

Reads a construction spec and returns the clauses that shift cost onto you, each one citing its page. Tested on real FIDIC and French tender documents.

Where I'm taking this

The scarce skill is not using AI. It is getting people who were never software developers to build real tools with it, inside organizations that were never software companies. I've done it for 200+ people. I'm turning that into work I do in the open: teaching Claude Code and applied AI to engineers, and helping firms in engineering and construction actually adopt it.