Back to the cover

The
Archive

A wider selection of this year's work, beyond the shortlist. Deep learning, generative AI, robotics, data engineering and symbolic AI, each with what I actually did on it.

Projects
06
Year
2026
  1. Illustration of the Unitree Go2 quadruped robot, three-quarter view.

    01 / Robotics / final-year project

    Quadruped Robots on Construction Sites

    Ongoing, with the innovation team at Bouygues Construction

    A use-case and feasibility analysis for Bouygues Construction: can a Unitree Go2 quadruped robot become a genuinely useful tool on their sites? YOLO object detection is in progress.

    • Python
    • Unitree SDK
    • YOLO

  2. 02 / Generative AI / hackathon

    HRévolution

    HrFlow.ai GenAI & HR Hackathon, team of 3

    Scores job applications continuously and explains every AI decision to the recruiter. Second place out of 50+ participants, demoed live to HR professionals.

    • React
    • FastAPI
    • LLMs
    • HrFlow.ai API
    • Docker
    Source
  3. An infant brain MRI slice, its ground-truth segmentation and the compact model’s prediction, side by side.

    03 / Deep learning / medical imaging

    Brain MRI Segmentation

    Solo personal project on the MICCAI iSeg-2017 challenge

    Segmenting infant brain MRI with a compact U-Net: 0.890 Dice with 95k parameters.

    • Python
    • PyTorch
    • U-Net
    Source
  4. staff decisionsSimulatormulti-agent, NumPyData pipelineParquet + DuckDBForecastingquantile boostingDetectionresiduals + rulesAllocationMILP (HiGHS)Orchestratorclosed loop, 15 min step

    04 / Simulation / digital twin

    Smart Festival

    EPITA SCIA semester project, team of 3

    An AI-driven digital twin of a festival: a multi-agent simulation, attendance forecasting, anomaly detection and closed-loop staff allocation. I led the design and wrote the final report.

    • Python
    • NumPy
    • DuckDB
    • Parquet
    • MILP (HiGHS)
  5. datalakeIoT sensorsmessage streamsIngestion APIgatewaymy partKafkatopicsSpark · bronzeraw, in Scalamy partSpark · silverincrementalmy partAlerts + analysisreal time, historical

    05 / Data engineering / streaming

    Real-time IoT Data Platform

    EPITA data engineering course, team of 4

    Sensor streams in, real-time alerts and historical analysis out. My part: the ingestion gateway publishing to Kafka, and the bronze and silver data-lake layers in Spark.

    • Scala
    • Apache Spark
    • Apache Kafka
    • sbt
  6. Puzzleunique solutioncommon interfaceBacktracking (MRV)Dancing LinksCP-SATSATGenetic algorithmBenchmarktimenodes exploredmemory

    06 / Symbolic AI / benchmark

    Sudoku Solver Benchmark

    EPITA symbolic AI course

    Five solving paradigms behind one interface (backtracking, Dancing Links, CP-SAT, SAT and a genetic algorithm), measured on the same puzzles for time, nodes explored and memory.

    • Python
    • OR-Tools CP-SAT
    • PySAT
    • Streamlit
    • pytest
    Source