AI/Data tech lead · Nantes and remote
Clément CharlesBook 30 min
AI/Data tech lead · LLMs in production · TypeScript

Your legacy, set straight.Your ideas, shipped to production.

I take over applications that grew too fast, without rewriting everything. And I build new ones: products, APIs, internal tools, AI in production.

// The LLM proposes, the schema disposes.

result: Order

  • client: "Menuiserie Leroux SARL"
  • reference: "CRT-450"
  • quantity: 12
  • delivery · to check: 2026-10-14
  • price: not provided

after the takeover

  • releases: every week, stress-free
  • incidents: caught before your customers
  • new features: without breaking things
  • the code: documented and tested
  • the team: at home in the codebase
50+sources plugged into the data/AI platform I lead
1developer across the whole chain, from collection to validated data
100,000+Android installs of the FilmoTV app
4major accounts delivered in agency: Google, Microsoft, Canal+, Club Med
01 · What I do

Set things straight. Build what’s next.

A hands-on tech lead: I design, I code, and I leave behind a codebase your team understands.

legacy01

Take over existing code without breaking it

Audit, gradual typing, tests around what already works, step-by-step refactoring: the application keeps running while it gets back on its feet.

Strict TypeScript · Tests · Step-by-step refactoring
back-end02

Build a product from scratch

From idea to production: clear architecture, APIs with versioned contracts, observability and CI/CD from day one.

Node.js · NestJS · PostgreSQL · Docker
data03

Your data flows, automated

Multi-source collection (APIs, scraping, e-mails, attachments) and ETL pipelines, down to the single format your tools need.

Playwright · ETL · PostgreSQL
llm04

AI where it actually helps

Invoices, PDFs, e-mails: LLM extraction held in check by a strict schema. What’s missing is flagged, never made up.

Claude · Mistral · Strict TypeScript
tools05

Your spreadsheets replaced by a real tool

Internal tools, dashboards and reporting, with a plain React interface built around the business.

Next.js · React · TypeScript
team06

A team that levels up

Structured code reviews, shared conventions, pairing and short documentation: the quality stays after I leave.

Reviews · Conventions · Pairing
02 · A closer look at your data flows

Four steps. No field made up.

Every document goes through the same chain, from the inbox to your tools. The one I run in production.

  1. 01

    Collect

    APIs, scraping, inboxes, attachments. More than 50 sources plugged into the platform I lead.

  2. 02

    Read

    An LLM (Claude, Mistral) spots the useful fields, even in a badly written e-mail or an attachment.

  3. 03

    Validate

    Every field goes through a strict TypeScript schema. What’s missing is flagged, never made up.

  4. 04

    Deliver

    A single format, ready for your tools: APIs, databases, dashboards.

03 · Case studies

Missions, not promises.

Case 01Today

Waste traceability

The problem

More than 50 data sources, each in its own format: APIs, websites, e-mails, attachments.

What I did

As the sole developer, I designed the whole platform: automated collection, LLM reading, validation against a strict schema.

The outcome

A single format, in production since February 2026.

Case 02Agency

Internal video tools for Microsoft

The problem

Existing internal video tools, with vulnerabilities reported by Microsoft’s security audits.

What I did

Took over the existing code, fixed the reported flaws, integrated Azure Media and DRM (PlayReady, Widevine).

The outcome

Tools brought back into compliance, without starting from scratch.

Case 03Agency

FilmoTV mobile app

The problem

A film app that had to play protected videos and download them for offline viewing.

What I did

Built the video player, the DRM protection and the downloads in the mobile app.

The outcome

More than 100,000 installs on Android.

Case 04Agency

Immersive room for Google

The problem

An immersive room where several screens had to react together, in real time.

What I did

Designed and built the internal application driving the connected screens: Node.js back end, React interface, WebSocket sync, GitLab CI/CD.

The outcome

In production at Google in Paris, London and Malaga.

04 · Method

Foundations that stay easy to maintain.

  1. 01

    Frame

    Quick scoping, architecture diagram, actionable backlog.

  2. 02

    Build

    Strict typing, code reviews, tests where they matter.

  3. 03

    Hand over

    Logs, traces and metrics from the start. Short documentation, knowledge transfer.

05 · Shall we talk?

Legacy to take over, or an idea to launch?

Thirty minutes to take stock together and leave with a clear plan.

// The LLM proposes, the schema disposes.