Thomas L.

Embedded Systems & IoT

Design and programming of ESP32 and STM32 microcontrollers (C/C++) to control firmness, resistance, and other physical measurement sensors.

Implementation of IoT protocols (Wi-Fi, Bluetooth LE) for real-time data transmission to field interfaces and cloud databases.

Setup of an end-to-end stack: sensor → microcontroller → IoT gateway → cloud → Web/mobile interface.

Integration, calibration, and standardization of high-precision sensors (force, pressure…), ensuring instrumental reliability ≥ ±0.5%.

Applications & User Interfaces

Development of control applications and analysis tools in WinDev and C#.

Design of mobile applications (Flutter & React) for Android, with UI/UX optimized according to Figma mockups.

Deployment and maintenance of responsive web interfaces for field data visualization.

Embedded AI & Cloud Computing

Integration of AI algorithms on microcontrollers for local processing and real-time anomaly detection.

Development of Python servers (Flask/FastAPI) and deep learning pipelines (PyTorch/TensorFlow) for predictive analysis and decision-making dashboards.

Industrialization & DevOps

Development of test and validation protocols (unit, integration) for each prototype before production.

Flagship Project: Northcopter B1 Drone

Electronic prototyping and sensor integration (CAN bus) for agro-environmental data collection.

Configuration of ArduCopter firmware and tuning of PID loops for optimized stability and autonomy.

Creation of a custom web platform for user service management in a Node.js environment (Express/EJS). Complete configuration of the network infrastructure: EOIP tunnels, security, VXLAN, segmentation, monitoring. Deployment and management of production servers (Linux, containers, high availability hosting). Design of the showcase website, SEO strategy, and management of targeted Google Ads campaigns.

Context:

Mirror404 offers jewelers an immersive shopping experience through virtual online try-ons. Integrated into their e-commerce platform, the module allows customers to digitally try on rings, necklaces, and earrings before purchasing.

AI Design & R&D

Feasibility study and architectural choice between GANs (StyleGAN2, CycleGAN) and diffusion models (DDPM) for photorealistic rendering on faces/wrists.

Design of training pipelines in Python (PyTorch) and Julia (Flux.jl) for image pre-processing (keypoint detection, fine segmentation).

Creation of manually annotated datasets (10,000+ images) and augmentation via style transfer techniques to enrich the diversity of jewelry and skin tones.

Training & Optimization

Deployment and management of GPU clusters (NVIDIA A100) under Slurm and Kubernetes, job orchestration with Docker and MLflow.

Front-end integration into e-commerce CMS (Shopify, Prestashop) via webhooks and JavaScript micro-widgets.

Technical documentation and training of DevOps teams for continuous deployment (GitLab CI/CD) and best practices in data science.