Case studies

Three public reference points for how I work

Technical software, AI prototypes and workflow automation. Protected details stay private, while the problem, contribution and decisions remain clear.

Software contribution at CETU — official reference

Tunnel-lighting software

Situation

Technical sizing software has to preserve domain rules, remain understandable and meet calculation and rendering constraints.

Contribution

I contributed to modernizing the software with C++, Qt and Python, around calculation, simulation, rendering and user experience.

Before rewriting, map what cannot break, measure critical paths and modernize progressively.

C++QtPythonEmbedded LinuxSimulation

View public reference

Public projects · GitHub

AI Models & My GPT-2

Situation

Understanding an AI model means connecting architecture, training, evaluation and observed limitations.

Contribution

These public repositories document Python and PyTorch experiments, including a GPT-2 implementation and training or evaluation pipelines.

A useful prototype exposes its assumptions, test sets and what still needs validation before business integration.

PythonPyTorchLLMsEvaluation

View public reference

Public project · GitHub

AREA

Situation

Workflows split across several tools become difficult to maintain when interfaces, data and automations diverge.

Contribution

The public project explores an automation application connected to web and mobile interfaces, with Python APIs and React components.

Value comes from clear contracts between tools, observability and a controlled automation scope.

FastAPIPythonNext.jsReact Native

View public reference

Have a project in mind?

Let’s identify the smallest useful first step.

Send me the context, the current bottleneck and your target date. I’ll reply with a concrete first proposal.

Talk about your project See selected work View one-page