Hello! I'm Tancrède, a computer scientist and robotics engineer-in-training. I've always been captivated by the process of turning complex ideas into tangible reality, a passion that took me from a First-Class Honours CS degree at McGill to my current MSc in Robotics at EPFL, where I'm focused on building the brains behind intelligent machines.
From programming autonomous robots to creating practical, robust software systems, or spending sleepless nights designing and developing games, I'm blending technical skill with creative expression in all kinds of projects. And even outside of projects I'm learning non-stop. You'll find me lost in the trenches of AI/ML research, exploring novel open-source projects, or binging the latest talks, blogs & videos on everything from control theory to systems design - chasing my curiosity down any rabbit hole I can find to stay on the cutting edge.
My goal is to build resilient, impactful systems, whether they're navigating the physical world or processing massive datasets. Looking ahead, I am eager to translate my academic background into industry experience and am actively seeking a challenging 6-month internship beginning in late Summer/Fall 2026. I'm particularly excited by opportunities in ML engineering, robotics, and applied AI research. I'm always eager to connect with others who share a fascination for this fast-moving field - whether you want to discuss the latest research, exchange ideas, or just say hi, please feel free to reach out.
A control project demonstrating autonomous rocket landing using Robust and Nonlinear Model Predictive Control (MPC). Engineered a 12-state NMPC for propulsive landing and Robust Tube MPC for vertical descent to guarantee collision avoidance under stochastic disturbances.
A decentralized auction system for heterogeneous swarms using market-based algorithms to coordinate and divide labor. Included a gossip protocol for local consensus and a multi-step task scheduler (TSP approximation) to optimize throughput under battery constraints in a Webots simulation.
An autonomous navigation stack for a Thymio robot featuring down-cam global navigation/planning, Braitenberg local avoidance, and EKF state estimation. Implemented a hierarchical controller to switch between navigation and obstacle avoidance behaviors, effectively managing static and dynamic obstacles.
An agentic framework for autonomous, semantic scene understanding. This project leverages LLMs to guide an agent through an interactive, simulated environment (ALFWorld) using text-based actions, gathering clues to infer the inhabitant's occupation via Bayesian-style belief updates.
A hands-on Unreal Engine 5 project where I learned C++ by building a small shooter to explore input handling, AI, and animation systems. Implemented behavior trees, state machines, and NavMesh-based pathfinding to prototype intelligent enemies and fluid character animations.
A Flask + SQLite single-page planner to let McGill students map courses across terms and track degree progress. Uses htmx for dynamic, server-driven UI updates and a scraped course dataset to simplify course selection and graduation tracking.
Hackathon web app for interview practice that records responses, transcribes speech with Whisper, and uses GPT-4 to generate feedback on clarity and professionalism. Includes webcam-based eye-contact tracking and metrics like speaking speed and word-density to give actionable practice insights.
Course project implementing a manager-worker distributed system using Apache Zookeeper to coordinate task distribution and worker availability. Built an availability-based load balancer and non-blocking manager threads to reliably assign tasks and collect results at scale.
.BhwRnRRF.png)
A geography quiz game pipeline that generates accurate mesh-based maps from Natural Earth data and builds interactive Unity quizzes. The project includes a multiprocessing Python mesh generator and tooling to author and share geography quizzes with precise GIS data.
A vectorized 2D Eulerian fluid solver implemented for a computer animation course, following Jos Stam’s real-time fluid method. Focused on stable advection, diffusion, projection steps and performance improvements via numpy vectorization.
 screenshot.B0_aY0o7.png)
A file system built from scratch in C, featuring a superblock, inodes with direct/indirect pointers, and a free-block bitmap. Implemented a full API for file operations like creating, writing, reading, and seeking, providing a low-level understanding of disk and data management.
A group data science project analyzing news coverage of 'Killers of the Flower Moon' against other major films. My contributions included scripting the data collection via NewsAPI, computing word frequency and tf-idf scores, and creating visualizations like stacked bar charts and a network graph to compare article topics and coverage proportions.
A top-down wave-based Unreal Engine game made to practise level design, AI blueprints, and gameplay loop mechanics. Players survive waves of enemies, collect pickups, and manage HUD elements while AI-driven enemies use NavMesh and blueprint logic to chase and attack.
A procedurally-generated 2D roguelike built in Unity with handcrafted level generation, A* pathfinding, and a persistent save system. Features increasing difficulty, a skill tree, and varied enemy behaviours to create replayable dungeon runs.
A Unity platformer where players outrun rising water and navigate hazards while collecting coins; includes collision/tilemap systems and reloadable weapons. Built to explore physics, collision layers, and responsive level design for platforming challenges.
Laboratory of Intelligent Systems @ EPFL Designing a lightweight deep learning pipeline for dense depth estimation on raw fisheye imagery. Targeting real-time inference on embedded drone hardware (NVIDIA Jetson) by bypassing standard rectification latency and artifacts.
McGill Robotics Refactored the Dockerized Unity simulation to improve code clarity and cut frame times by ~50%. Contributed to a critical stack migration from ROS 1 to ROS 2 and led a documentation overhaul to accelerate developer onboarding.
Investigated Generative Flow Matching models, specifically Acceleration Flow Matching (AFM). Extended a PyTorch codebase to benchmark AFM against standard baselines, analyzing performance gains in modeling higher-order dynamics.
EPFL I'm specializing in Mobile Robotics, with a focus on Embodied AI. After knocking out the rigorous mathematical foundations (MPC, Convex Optimization), I'm now applying them to physical systems - currently doing research on real-time vision for drones, alongside coursework in RL, CV and deep learning for autonomous vehicles.
McGill University I graduated with First-Class Honours (3.85 GPA), dipping my toes in various advanced topics like Machine Learning, Computer Animation, Computational Biology, and Distributed Systems. Outside of class, I was heavily involved in McGill Robotics, GameDev McGill, and competed in hackathons like McHacks and PharmaHacks.
Epsom College My secondary education in the UK, where I earned Art and Computer Science awards, laid the groundwork for my technical and creative interests. Beyond academics, I was active in the school's tennis and field hockey teams, and completed my Duke of Edinburgh's Award.
