Skip to content
View ameroyer's full-sized avatar
🥐
🥐

Block or report ameroyer

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ameroyer/README.md

Amélie Royer  🥐

Deep Learning Research Engineer · Kyutai [kjutaɪ] · Paris, France

Building open-science AI at kyutai.org 🇫🇷 · Previously Qualcomm AI Research 🇳🇱 · PhD from IST Austria (2020) 🇦🇹

Website Twitter Google Scholar


🔬 Research

My work sits at the intersection of computer vision, multimodal learning, and neural network efficiency. I care about building models that are both powerful and practical. I'm very interested in learning from multiple modalities and how to best make use of raw data that is inherently multimodal.

Project Higlights Venue What it is about
🏠 CASA -- Cross-Attention Strikes Back ECCV 2026 Vision-language model with cross-attention for scalable streaming inference
M👁️shiVis -- Kyutai with an "eye" CVPR 2026 Adds visual understanding to the Moshi speech model with a data-efficient training pipeline
MSViT -- Mixed-Scale Tokenization ICCV NViT Workshop, 2023 Dynamic token scaling for Vision Transformers based on image content
Scalarization for Multi-Task Learning NeurIPS, 2023 Large-scale study of multi-task/domain training dynamics + population-based optimization
Knowledge Distillation: A good teacher is patient and consistent (oral) CVPR, 2022 Patient & consistent teacher = surprisingly strong distillation recipe

→ Full list on my website


📓 Open-Source / Tools

I enjoy building tools for automating my daily tasks; Only one constraint: They need to have cute names and/or aesthetics.

  • 🍬 PeperNoten -- A CLI tool to parse arXiv papers into Obsidian-ready markdown research notes. It slowly builds a knowledge base of research papers, clustered into topics, which you can then connect to via MCP.
  • 📖 Grimoire -- A research logbook built by parsing and comparing MLFlow runs: Select the runs you want to compare, Claude writes a first analysis for you and you can then add your own comments and notes. Like PeperNoten, you can access your grimoire via MCP to quickly check your previous conclusions and analyses.
  • 🍫 PRaline -- Did your number of PRs to review dramatically increase with the advent of Claude ? PRaline helps you get these first PR review comments in, and builds a knowledge base of common issues, bugs, pain points with each merged PR in your repo
  • 📎 PocketPal -- Bring some fun to your remote working days by turning Claude into a cute voice assistant using on-device Speech-to-Text and Text-to-Speech.

📓 Side Projects & Snippets

Small experiments and one-off builds:

  • 🎄 Advent of Code -- A list of my past Advent of Code participations
  • 🕵️ Codenames Solo -- solo mode for the Codenames boardgame, powered by GPT and Streamlit
  • 🌊 Glow JAX -- clean JAX/Flax implementation of the Glow generative model

🛠 Tech stack

Python JAX PyTorch Jupyter

Pinned Loading

  1. ameroyer.github.io ameroyer.github.io Public

    Personal websites (CV, publications and blog posts)

    Jupyter Notebook 3

  2. kyutai-labs/moshivis kyutai-labs/moshivis Public

    Kyutai with an "eye"

    Python 254 33

  3. kyutai-labs/casa kyutai-labs/casa Public

    A vision-language model with an improved cross-attention mechanism for scalable streaming inference

    Python 30 3

  4. glow_jax glow_jax Public

    An implementation of the Glow generative model in jax and flax

    Jupyter Notebook 8 1

  5. advent_of_code_2023 advent_of_code_2023 Public

    Advent of Code 2023 - Python solutions

    Jupyter Notebook

  6. pepernoten pepernoten Public

    Automated paper reading notes and growing knowledge base

    Python