Final-year Engineering student at UPSSITECH – University of Toulouse Former Research Intern at AIST, Japan 🇯🇵
Looking for a 6-month internship · March → August 2027
I'm a final-year engineering student specializing in Robotics and Interactive Systems at UPSSITECH.
I enjoy building complete intelligent systems at the intersection of robotics, computer vision, artificial intelligence and software engineering — from perception and autonomous navigation to robot manipulation and applied machine learning.
My recent work includes research on 3D perception and vision-language robotic manipulation at AIST Japan, autonomous Formula Student path planning with TLSe Racing, and an independent reproduction of EarlyEval for predicting AI agent outcomes from partial execution trajectories.
📍 Toulouse, France 🎯 Seeking an internship from March to August 2027
Computer Vision Research Team · Artificial Intelligence Research Center (AIRC) Apr 2026 – Aug 2026 · Tsukuba, Japan
Extended SVLR (Scalable Visual Language Robotics) for long-horizon robotic manipulation.
- Developed an open-vocabulary RGB-D 3D perception pipeline using Grounding DINO and SAM 2.
- Designed persistent world memory for object identity, visibility, spatial relations and pose history.
- Integrated perception, planning and execution with ROS 2, RMBench/RoboTwin and a real Franka Panda.
- Evaluated the complete system over 900 simulation episodes across 9 tasks, achieving 71.1% overall success, with six tasks reaching 92–100%.
Worked on autonomous trajectory planning and vehicle control for a Formula Student race car.
- Developed and evaluated RRT*-based trajectory planning strategies.
- Implemented trajectory smoothing and curvature-aware vehicle control.
- Integrated the final approach into a ROS 2 / Gazebo Formula Student simulation.
- Reduced simulated lap time from 44 s to 30 s (-28.6%).
Path Planning · Motion Planning · 3D Perception · Robot Manipulation · Robot Kinematics
Machine Learning · Deep Learning · Computer Vision · Vision-Language Models · LLMs
| Project | What I built | Main technologies |
|---|---|---|
| 🏎️ Path Planning Studio | Interactive 3D environment for prototyping autonomous racing algorithms before ROS 2 integration. RRT*, QP trajectory optimization, vehicle simulation and real-time telemetry. | TypeScript · React · Three.js · RRT* · QP |
| 🧪 EarlyEval Reproduction | Independent reduced reproduction of EarlyEval for predicting AI agent outcomes from partial execution trajectories. Evaluated 700 trajectories across 350 SWE-bench Verified tasks using behavioral and TF-IDF/SVD features, calibrated LightGBM models and held-out-agent evaluation. | Python · LightGBM · scikit-learn · TF-IDF · SVD |
| 🚨 Incident Co-Pilot | AI-powered incident management platform with contextual assistance, vector similarity search and incident summarization. | TypeScript · NestJS · Next.js · FastAPI · pgvector · Docker |
| 📊 Churn Prediction Platform | End-to-end machine learning pipeline for customer churn prediction, reaching ~0.84 ROC AUC, with model comparison and an interactive analytics dashboard. | Python · Scikit-learn · XGBoost · PyTorch · Streamlit |
| 🦾 UR3 Kinematics & Trajectory Generation | Forward/inverse kinematics, Jacobian-based differential models and Cartesian-to-joint trajectory generation for a 6-DOF UR3 manipulator. | Python · NumPy · SymPy · Matplotlib |
| 🤖 Autonomous Mobile Robot | Academic mobile robot integrating autonomous navigation, obstacle avoidance, object tracking, voice control and a web HMI. | Python · C++/Arduino · OpenCV · Flask · Embedded Systems |
| 🧠 MobileNetV2 – CIFAR-100 | Two-stage transfer-learning and fine-tuning pipeline for CIFAR-100, reaching >71% test accuracy. | TensorFlow · MobileNetV2 · Transfer Learning |
Robotics Software · Computer Vision · Autonomous Systems · Robot Perception
AI / Machine Learning · Vision-Language Robotics · Intelligent Software Systems
