Lucien Rivière
  • HomeGo to
  • ProfileExperience, skills and CVGo to
  • MethodHow I work, layer by layerGo to
  • ContactSay helloGo to
  • ProjectsEverything I built, filterableGo to
  • CollaboratorsEveryone I built withGo to
  • SignalMapA mobile coverage map of Réunion and Mayotte: click anywhere and see the signal you would get from about 3,100 antennas.Projects
  • ZappyA network survival game in three programs: a C++ server, a 3D viewer and AI players, with LLM role-playing agents as a bonus.Projects
  • MaskGeneratorA small neural network that finds the road lines in a camera image, and a 2D raycast that turns them into lidar-like distances to drive a car.Projects
  • CockerA prototype platform to design an app as a graph of services and deploy it on Kubernetes: sandbox to try it, production to expose it.Projects
  • RaytracerA C++ raytracer where everything (shapes, materials, lights, output) is a plugin, sped up with a BVH, with a live preview window.Projects
  • RoboCarA small real car that drives itself around a track: a camera, a lidar-like view of the road lines, and a model that learned from human driving.Projects
  • CVRIETwo machine-learning studies using only scikit-learn: classifying chest X-rays, and grouping patient testimonies into symptom clusters with no labels.Projects
  • MyTeamsA Teams-like chat in Rust: a multi-client server, a command-line client, a text protocol written as an RFC, and data that survives restarts.Projects
  • ArcadeA retro game console in C++ where games and screens are plug-in libraries: swap the game or the graphics (terminal, SDL, SFML) while it runs.Projects
  • Masked DeliveryA game made in a weekend for Global Game Jam 2026: pick the right mask for each caller, then survive the drive, with penalties that stack up.Projects
  • 42shA Unix shell written in C, tcsh-style: pipes, logic operators, globbing, aliases, variables, line editing and history, checked against Bash.Projects
  • CorewarA virtual machine in C where small programs fight inside a shared memory, with my own warrior and a 3D viewer in the bonus.Projects
  • A-DasHboarDA dashboard of widgets from third-party services, built on JustBetterAuth, an authentication library written from scratch.Projects
  • StoneAnalysisA Python tool that finds the notes in a sound file with a hand-written Fourier transform, and hides secret messages inside the audio.Projects
  • PipetoA security audit of a nuclear-reactor console full of planted vulnerabilities: find them, rank them, patch them and prove the fixes with tests.Projects
  • TardisCleaning SNCF train-delay data, training a model to predict delays, and a Streamlit dashboard. The model barely beat guessing, and the page says so.Projects
  • InterStonarThrow a rock into a simulated universe: planets that attract each other and merge, or shapes it hits, found by stepping with signed distances.Projects
  • SensAIA small chat program for a local AI model that talks to it with plain HTTP, and can force its answers into a JSON shape you define.Projects
  • This portfolioThe site you are on: one set of data, six themes. Each theme is a different way to explore the same work, from a familiar page to a game.Projects
  • Portfolio toysSmall interactive toys from this portfolio: Fourier epicycles, a themed crossword and a bot that walks where you click.Projects
  • snake_pyA contribution to Infamous97440's repository: 2 commits from me.Projects
  • MyTopA clone of the Linux top command in C and ncurses: live CPU, memory, swap and per-process usage read from /proc.Projects
  • Switch to CorporateThemes
  • Switch to SpaceThemes
  • Switch to Cocktail PartyThemes
  • Switch to GroveThemes
  • Switch to Voxel RedstoneThemes
  • Search the whole siteActions
  • Download CV (PDF)Actions
  • Copy my email addressActions
  • Turn the home page bot on or offActions
All themes

Tardis

Lucien Rivière

  • 14 commits
  • archived

About this project

Apr 2025 – May 2025 · Data cleaning, model and dashboard · team of 3

Cleaning SNCF train-delay data, training a model to predict delays, and a Streamlit dashboard. The model barely beat guessing, and the page says so.

Can we predict how late a French train will arrive? Tardis takes the SNCF data on train punctuality, cleans it, trains a model, and puts the result in a small web app where anyone can browse the statistics or try a prediction.

How it works

Most of the work is cleaning: broken lines, numbers stored as text, impossible dates, and station names with typos (Dinkerque for Dunkerque). Each repair is documented step by step. A model then learns from the clean data, and the dashboard shows delays by station and by hour, with a form to ask for a predicted delay.

The result is reported honestly: the model did not beat simply answering with the average delay. The notebook keeps that finding instead of hiding it.

What I took from it

  • A model that fails to beat the average is still a finding: it says the inputs are not the story, or that the preparation needs another look.
  • Cleaning data is most of the job, and it deserves to be written down as carefully as the model.

Cast

  • Samuel Jaglalesleepwalkers dev — where dreams become production code. Coding my dreams into reality.GitHubWebsiteLinkedIn
  • Lucas PlanterouxGitHub