01 / RESEARCHER
LOUPe
Questions, sources and the wider picture.
AYOUB OULAD ALI / CODE
Ayoub Oulad Ali — AI & Applied Mathematics projects
This page curates public repositories from GitHub @ayouboa30, prioritizing Generative AI, machine learning, representation learning, mathematical modeling and scientific computing over an exhaustive feed.
LOUPe beta 0.1.4 / LOCAL-FIRST WORKSPACE
Three agents.
One shared loop.
LOUPe is an experimental desktop workspace where specialized agents propose, challenge and write together — with the conversation context shared across the loop.
Available for Windows and Linux. The Windows installer prepares the local stack — WebView2, Ollama, qwen3:1.7b-flash, Node.js/npm, Codex, Claude Code and OpenCode — so the first run feels like a launch sequence, not a dependency hunt. The Linux build ships as a self-contained tarball.
0.1.4 adds a literature review that answers a research question with numbered, clickable citations — and names what the retrieved sources leave unanswered. It also fixes two visible interface bugs and brings text contrast up to WCAG AA. The installer stays at 118 MB, down from 332 MB before 0.1.3.
SHA-256 (Windows) 0C29A34B44F588347CF4DEDA197915703CFA2811D3444CE7A2481BEF21D4E2DE
SHA-256 (Linux) 42a942d09cbe3d1197218badd7531dddd5158a892cbdc0c71a763aae0b16c50e
Read the beta release notes ↗
02 / MATHEMATICS
MATh
Proofs, structure and careful reasoning.
03 / CODE
CODy
Implementations, tests and edge cases.
CONTRIBUTORS / LICENSE
LOUPe is source-available and open to contributors: ideas, issues, fixes and forks are welcome. It is released under the LOUPe Non-Commercial Source License 1.0.
Commercialization is prohibited. Personal, educational, research and community use are allowed; selling, reselling, paid hosting, paid services and commercial redistribution require written permission. This custom license is not OSI-approved.
Selected GitHub repositories
The cards below are loaded from the GitHub public API and ranked for relevance to my research profile. If the API is temporarily unavailable, the direct GitHub profile remains crawlable and accessible.
Research context
For the scientific context behind these projects, see the research page, which covers generative modeling, latent representations, physics-informed learning, uncertainty and medical imaging.