Joseph RIGAL — Robotics & AI Engineer / Athlete

01

About

I'm a final-year engineering student at École Centrale de Lyon, specialising in automation, physical AI and robotics.

My work lives at the intersection of data-science and hardware: giving robots the ability to understand a scene, act on it, and do so safely. Recently I won Paris Builds (a Y Combinator × Unaite hackathon), integrated on non-visual robotic perception at Wormsensing, and sped up a document-AI pipeline 10× at Probayes.

Off the screen, I'm a competitive middle-distance runner. In my first competitive season this year I've set personal records in every event from the 800 m to the 10 km and qualified for the French Cross-Country Championship, the same discipline, pointed at a different finish line.

Education

M2 Automation · Gen. Engineering

École Centrale de Lyon · 2021–2026

Focus

Robotics · Physical AI · CV

LeRobot · SO-101 · OpenARM

Athletics

800m → 10km · FFA licensed

Team Illegiteam · coach J. Ghibaudo

Latest

🏆 Paris Builds Hackathon winner

YC interview earned · 2026

Python PyTorch LeRobot Computer Vision Transformers ROS Control Systems C / C++

02

Experience

Where I've studied, built and taught.

Work

April 2026 — Present

R&D Robotics Intern

Wormsensing

Pioneered a proof of concept for non-visual robotic perception: integrated Wormsensing's Dragonfly® vibration sensor into LeRobot and routed live spectrograms as "hacked camera" inputs, training robots to identify objects hidden inside opaque boxes, tasks impossible with cameras.

Deep dive into tactile sensors: Dragonfly® (piezo), accelerometers, cell-force, strain gauge ...
Hands-on analysis of state-of-the-art robot-learning models

May 2025 — August 2025

Data Scientist Intern

Probayes

Engineered a 10× speed-up in PDF-to-Markdown conversion for DocParser by profiling and optimising the critical bottlenecks in a deep-learning document-layout-analysis pipeline.

PDF to MD pipeline optimisation
Transformer-based document layout training and optimizing

September 2024 — Present

Mathematics Tutor

Institut Fibonacci

Private mathematics lessons for undergraduate students.

Explaining complex concepts clearly

Education

2024 — 2026

Master's in Automation

École Centrale de Lyon

Advanced studies in control systems, robotics and physical AI.

2023 — 2026

General Engineering · CS Specialisation

École Centrale de Lyon

Core engineering curriculum (mechanics, fluid dynamics) with a specialisation in machine learning, computer vision and artificial intelligence.

2021 — 2023

Preparatory Class · Math

Lycée Champollion

Deep dive into advanced fundamental mathematics and physics.

03

Projects

Robotics, perception and safety.

// 001 🏆 Winner — Paris Builds (YC × Unaite)

VIGIE — Robot Safety Watchdog

36h hackathon · won a Y Combinator interview

An external runtime safety layer for learning-based robots. VIGIE watches the robot and its workspace through environment cameras, detects dangerous situations, and can pause, hold and resume any robots run when the scene becomes unsafe. Built because robot safety shouldn't depend on the policy alone. Built in 36 hours with a live demo harness interfaced with LeRobot and OpenArm.

Computer Vision Robot Safety LeRobot OpenARM Real-time

Code The story

// 002 R&D

Non-Visual Perception

Wormsensing · LeSpectrobot

Unlocking robotic perception without cameras: a "dragonfly" vibration sensor integrated into LeRobot, with live spectrograms routed as "hacked camera" inputs. Comparing tactile sensors solutions: dragonfly, accelerometer, load cells, strain gauge ...

LeRobot Spectrograms Sensor integrations VLA

Code

// 003 Robotics

LeJunior — Vision Pick & Place

SO-101 · YOLO11 · ACT policy

A vision-guided pick-and-place pipeline: commanding a VLA throught visual input and heuristics to fill boxes of different capacities (for instance, 15 × 40 units).

SO-101 YOLO11 ACT Policy

Code

// 004 In progress

SO-101 Electromagnet Mod

Custom end-effector

Extending the LeRobot library with an electromagnet end-effector to customise the SO-101 arm. Adding magnetic grasping to the toolkit and integrating it cleanly into Lerobot control stack.

LeRobot Hardware Electro-magnet

// 005 Planned

Autonomous Bin-Picker Rover

Mobile manipulation · ROS

A mobile rover paired with an SO-101 arm to autonomously find and pick up clutter around my room, a hands-on path into ROS, navigation and full-stack mobile manipulation.

ROS Mobile Robotics Manipulation Autonomy

04

Track & Field

Personal records, from the 800m to the 10km road.

800 m1:56.65

1500 m3:52.69

3000 m8:51.50

5 km15:08

10 km30:45

Highlights

165th at the French Cross-Country Championship (2026)
Dropped my 10 km from 32:02 to 30:45 (+1:17 in eight months)
1500 m from 3:59 to 3:52 in a single year (ranked 336 overall in the 2026 season)
5 km debut in 15:08

Season Timeline

May 2026 — Present

Outdoor season

Ran a 3:52 1500 m after a few setbacks; still chasing a clean 800 m through variable wind and form.

Jan 2026 — May 2026

Cross-country season

First-ever cross season, I qualified for the French Championship in my first year, finishing 162 out of 352 qualified.

Sep 2025 — Jan 2026

Professional training

Structured training and recovery with coach Jérémy Ghibaudo at Team Illegiteam. Ran a 30:45 10 km in Vénissieux before cross season.

Apr — Jul 2025

First track season

Set every record from 800 m to 10 km road, most on a single race per distance.

2024 — early 2025

Getting into running

Discovered road racing with a 33:47 10 km, no formal training. Got my FFA license, qualified for the French 10 km Championship (85th).

05

Contact

Let's build together,
new ideas?

Open to robotics, AI roles, collaborations and hard problems. Reach out, I'm looking forward hearing your ideas.

joseph.rigal@etu.ec-lyon.fr