Skip to content
View Just1truc's full-sized avatar
🐢
Slowly but efficiently learning
🐢
Slowly but efficiently learning
  • Epitech
  • Lyon

Block or report Just1truc

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.

Please don't include any personal information such as legal names or email addresses. Maximum 100 characters, markdown supported. This note will be visible to only you.
Report abuse

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

Report abuse
Just1truc/README.MD

👋 Justin Duc (Just1truc)

“Attention is all I need”
Building performant AI infrastructure from first principles.


🧑‍💻 About Me

I’m a final-year student at EPITECH Lyon and recently completed an exchange year at Tsinghua University. While at Tsinghua, I was part of the team that earned the Leader Award at Chengdu 80 (2024) — a prestigious international FinTech competition. With 5 years of hands-on software experience across the stack, I’ve explored nearly every major domain of computer science and chosen to specialize in Artificial Intelligence.
I enjoy digging into the math and systems engineering that power modern ML—then open-sourcing the tools I wish existed.


💼 Core Competencies

Field Keywords
AI / ML sub-quadratic attention, autograd engines, model compression, NLP
Systems C / C++, GPU & SIMD optimization, multithreading, low-level profiling
Data Infra distributed training, big-data ETL, high-throughput I/O
Languages Python, C, C++, TypeScript, JavaScript, Bash, SQL, …
Frameworks PyTorch, CUDA, React/Next, Docker, WebRTC

🚀 Notable Projects

Project Brief
treat Novel sub-quadratic attention framework built on Binary-Tree Attention Trees composed from Linear Attention. Achieves O(n log n) context aggregation while retaining full-sequence fidelity.
saute Multi-party discourse processor with hierarchical EDU-level parsing and speaker-aware memory, implemented via stacked linear-attention blocks for efficient long-context reasoning.
dtorch Minimal autograd + neural-network library—clean codebase for educational deep-learning research.
Argos Real-time remote desktop manager leveraging WebRTC, Socket.IO, and low-level thread orchestration.
excalidraw-latex-gen CLI to generate LaTeX-based Excalidraw diagrams automatically.
Abricot-Norminette Ultra-fast, multithreaded C style-checker adopted by EPITECH cohorts.

📊 GitHub Stats

Top Languages GitHub Stats


🌐 Where to Find Me


Although I’ve delved into everything from compiler design to distributed systems, AI remains my primary focus. Blockchain is about the only field I haven’t explored—yet.

Thanks for stopping by! 🚀

Pinned Loading

  1. tzhengtek/saute tzhengtek/saute Public

    SAUTE is a lightweight transformer-based architecture adapted for dialog modeling

    Python 2

  2. treat treat Public

    Research framework for sub-quadratic attention using LA and Knowledge Distillation

    Python 2

  3. Abricot-Norminette Abricot-Norminette Public

    Abricot Norminette is meant for EPITECH Projects. [ PROMO 2026 ]

    Python 47 8

  4. dtorch dtorch Public

    A Package created by a student for IA Model creation and saving that reproduce pytorch module and autograd system.

    Python 12

  5. Argos Argos Public

    Can see everything, beware of its omniscience, kneel before its greatness.

    TypeScript 2

  6. excalidraw-latex-gen excalidraw-latex-gen Public

    A browser extension to automaticaly generate latex on excalidraw

    JavaScript 6 1