Founding Computer Vision Engineer at Viseio · Toulouse

Maelys Risset

I build computer vision systems that run in production, in real time: from classical image processing to Deep Learning, from the algorithm all the way to on-site installation.

Get in touch See my background
ENSIIE engineer
& Master's in AI, Paris-Saclay
From algorithm to production
design, optimisation, deployment
English C1
international client relations

What I do

Industrial vision

Defect detection and segmentation on multispectral images: mathematical morphology, filtering, difference of Gaussians, edge detection. Fast and interpretable.

Deep Learning

Classification and segmentation (MobileNet, EfficientNet, UNet, Transformers), from building the dataset to evaluating the models.

Real-time inference

Production deployment under throughput constraints: ONNX, OpenVINO, CUDA, quantisation, batching, asynchronous pipelines.

Experience

  1. Since Sept 2026

    Founding Computer Vision Engineer

    Viseio

    Part of the founding team, building the computer vision components of the product.

  2. 2024 — 2026

    Image Processing & AI Engineer

    MAF RODA (MAF Group) · Montauban, then Perpignan, France

    MAF RODA designs grading machines that sort fruit and vegetables in packing houses. I developed the optical sorting software: multispectral images of each fruit are analysed in real time to grade it by colour, shape, size and defects.

    • End-to-end pipeline, from image acquisition (Matrox Imaging Library) to the sorting decision, in C# / .NET.
    • Defect detection algorithms: classical image processing first, Deep Learning where it reaches its limits.
    • Training PyTorch models (MobileNet, EfficientNet, UNet) and building the datasets.
    • Optimisation to keep up with line throughput: quantisation, pruning, batching, asynchronous pipeline, GPU offloading.
    • Tooling for labelling, training and evaluating models.
    • International client relations: demos, specifications, on-site installations, support.
    • C#
    • .NET
    • Matrox MIL
    • Python
    • PyTorch
    • ONNX
    • OpenVINO
    • CUDA

  3. 2023 · 7 months

    Deep Learning Engineer (research internship)

    INRIA, SERPICO team · Rennes, France

    Unsupervised motion segmentation of small objects in aerial image sequences. Integrated a Slot Attention mechanism into a Transformer-UNet architecture, generated synthetic data and added prior knowledge about object size and shape.

    • PyTorch Lightning
    • Weights & Biases
    • Docker
    • Singularity
    • Linux

  4. 2022 · 4 months

    Machine Learning Engineer (research internship)

    LSSI · Trois-Rivières, Québec, Canada

    Sound source localisation in adverse acoustic conditions: regression CNNs on time-frequency images from a wavelet decomposition, noise robustness study (Gaussian, pink, babble) and data augmentation.

    • MATLAB
    • CNN
    • Wavelets

  5. 2021 · 3 months

    Full-stack Developer (internship)

    KUSO · Lyon, France

    Analytics module for a digital marketing app aimed at real-estate agencies: automated data collection through the Facebook Graph API and interactive charts.

    • C#
    • .NET
    • React
    • TypeScript
    • PostgreSQL

Projects

Personal projects

Android apps I design and build on my own with Expo, React Native and TypeScript. They work offline: no account, and data stays on the phone (SQLite).

Mobile app

Kesako — flashcard revision

Imports questions and answers (text, spreadsheet or Anki export) and derives seven exercise types from them: multiple choice, matching, fill-in-the-blank, ordering… Tracks mastery of each card, with optional card generation through the Claude API. Over 1,100 unit tests.

  • React Native
  • TypeScript
  • SQLite
  • Claude API

Mobile app

Cehou — geography quiz

Interactive map quizzes: countries, flags, capitals, cities, oceans and seas, in four game modes. Reproducible data pipeline (Natural Earth, Wikidata), precomputed projections and Skia rendering that stays sharp at any zoom level.

  • React Native
  • Skia
  • d3-geo
  • Wikidata

Mobile app

Sabrul — workout sessions

Build and play workout sessions: a library of 294 illustrated exercises, filterable by muscle and equipment, and a full-screen player with timer, vibrations and voice announcements.

  • React Native
  • TypeScript
  • SQLite
  • Python

Mobile app

AppRecipe — recipes & fridge

Suggests what to cook based on what's in the fridge, allergies, available time and cravings, using a scoring algorithm. Shopping list grouped by aisle and nutrition facts (CIQUAL table).

  • React Native
  • TypeScript
  • SQLite

Web app · since 2022

FamilyNest — shared home booking

Manages homes shared between family members or friends: stay requests, per-home calendar and booking conflict detection. JWT authentication and real-time notifications over WebSocket.

  • MongoDB
  • Express
  • React
  • Node.js
  • WebSocket

School projects (ENSIIE)

Multi-agent systems — Warbot

Decentralised strategy for robot teams with only local perception: coalitions, intercepting enemy messages.

  • Processing

3D rendering engine

Rasteriser from scratch: geometric primitives, camera projection, SDL display.

  • C++
  • SDL

Minimax & alpha-beta

AI for Nim and Connect Four, playable against the computer or AI versus AI.

  • OCaml

Formal methods — medical records

Event-B model of access rights to guarantee the confidentiality of patient data.

  • Event-B
  • Rodin

SCRUM web project — NASA API

NASA's daily images, user accounts and favourites, built in sprints with rotating roles.

  • React
  • Laravel
  • SQL

Fish school simulation

Reynolds' flocking model: attraction, alignment and separation between agents.

  • NetLogo

Skills

Image processing
Mathematical morphology, spatial and frequency filtering (FFT), denoising, edge detection, wavelets, Matrox Imaging Library
Deep Learning
PyTorch, PyTorch Lightning, TensorFlow · CNN (ResNet, MobileNet, EfficientNet, UNet, YOLO), Transformers, LSTM, Slot Attention · Weights & Biases
Deployment
ONNX, OpenVINO, CUDA · quantisation, pruning, reduced precision, batching, parallelisation
Languages
Python, C#, C++, C, Java, TypeScript, JavaScript, Kotlin, MATLAB, R, OCaml, PHP
Web & data
.NET, React, Node.js, Express, Laravel · PostgreSQL, MySQL, MongoDB · NLP (SpaCy, LLMs)
Tools
Git, Docker, Singularity, Postman, Qt, LaTeX · Windows, Linux · SCRUM

Education & languages

Education

  • Engineering degree in computer scienceENSIIE · 2023
  • Master's in Artificial IntelligenceUniversité Paris-Saclay · 2023
  • Preparatory classes (CPGE)Lycée du Parc · 2018 – 2020

Languages

  • FrenchNative
  • EnglishC1 · Linguaskill Business (Cambridge), daily professional use
  • GermanIntermediate

Outside work

Contact

A question, a project, or just want to talk computer vision? Drop me a line.