Sleeping
RL
1
Code Security Review OpenEnv
๐ก
Scan code to find security vulnerabilities
Learning Framework
Building AI agents and tools that work in the real world โ from code security review to adaptive education.
Inmodel Labs is a small non-profit team researching and building open AI systems. Our work sits at the intersection of agentic AI, application security, and adaptive learning. We believe the best way to advance AI is to put it inside hard, real-world environments โ and measure what actually happens.
<div class="project-card purple">
<div class="project-icon">๐ก๏ธ</div>
<div>
<a class="project-name" href="https://huggingface.co/inmodel/open-env">
OpenEnv โ Code Security Review
</a>
<p class="project-desc">
A reinforcement learning environment for training AI agents on real-world AppSec triage.
Agents discover hidden source files and identify vulnerabilities across three difficulty levels.
Built for the <strong style="color:#c4b5fd;">Meta PyTorch OpenEnv Hackathon</strong>.
</p>
<div class="tag-row">
<span class="tag">RL Environment</span>
<span class="tag">Security</span>
<span class="tag">Python</span>
<span class="tag">JavaScript</span>
</div>
</div>
</div>
<div class="project-card green">
<div class="project-icon">๐</div>
<div>
<div class="project-name">EduAI โ Adaptive Learning Platform</div>
<p class="project-desc">
An AI-powered ed-tech platform for students aged 14โ22. Handwritten test review via OCR,
AI tutor chatbot, flashcard generation, personalized study schedules, live classes,
and gamification to keep learners engaged.
</p>
<div class="tag-row">
<span class="tag">Ed-Tech</span>
<span class="tag">Agentic AI</span>
<span class="tag">Next.js</span>
<span class="tag">TypeScript</span>
</div>
</div>
</div>
</div>
Evaluated with meta-llama/Llama-3.3-70B-Instruct
| Task | Difficulty | Score |
|---|---|---|
python-off-by-one |
Easy | 0.883 |
js-idor-auth |
Medium | 0.500 |
python-pickle-deserialization |
Hard | 0.512 |