A student-driven research lab dedicated to building, breaking, and experimenting with AI — from first-principles implementations to original research.
RLS (Research Lab SUP'COM) is a research-first community for students who want to go beyond coursework and consumption — reading a paper is the starting point, not the destination. We reimplement architectures from scratch, run our own experiments, question results, and push toward original contributions.
We are not a club that talks about AI. We are a lab that builds it.
Our mission is simple: give students a real research environment — one with reproducible code, technical rigor, and the mentorship needed to go from "I read this paper" to "I can rebuild, question, and extend it."
| Area | Focus |
|---|---|
| 🧠 Machine Learning & Deep Learning | Foundations, architectures, and training dynamics |
| 👁️ Computer Vision | Perception, representation learning, visual reasoning |
| 🎮 Reinforcement Learning | Decision-making, control, multi-agent systems |
| 🌍 World Models | Learned simulators, planning, model-based agents |
| 🎨 Generative AI | Diffusion, autoregressive, and generative modeling |
| 💬 LLMs & Efficient Inference | Training, fine-tuning, quantization, serving at scale |
| ⚙️ AI Systems / Research Engineering | Infra, tooling, and pipelines that make research reproducible |
Computer Vision |
Reinforcement Learning |
World Models |
Generative AI |
LLMs & Efficient Inference |
AI Systems |
- 🔬 Paper reimplementations — rebuilding landmark papers from scratch to understand them, not just cite them
- 🧪 Experiments — ablations, ideas that might not work, and the ones that do
- 📊 Benchmarks — evaluating models and methods against reproducible baselines
- 🛠️ Reproducible pipelines — training/eval code others can actually run
- 📄 Technical reports — write-ups of findings, negative results included
- 🚀 Original research — pushing toward novel contributions, not just replication
Learn → Build → Experiment → Research → Share
- Learn — study the fundamentals and the state of the art, deeply
- Build — reimplement it yourself; understanding comes from the code, not the abstract
- Experiment — test ideas, break things, iterate fast
- Research — turn solid experiments into original questions and contributions
- Share — publish reports, code, and findings so others can build on our work
RLS is built by students, for students — guided by mentors who push us further.
Members
List of current members — coming soon.
Mentors / Advisors
List of mentors and advisors — coming soon.
RLS is open to SUP'COM students who want to move from studying AI to building it.
- Join the lab — reach out through our contact channels below
- Propose a project — bring a paper, an idea, or a problem worth exploring
- Contribute code — help implement, benchmark, or document ongoing projects
- Attend sessions — paper discussions, implementation workshops, research talks
RLS — Research Lab SUP'COM From papers to prototypes.
