I am interested in annotation systems and deeply engaged with Transformer architectures and their optimization techniques. I am particularly fascinated by quantization methods for reducing model storage requirements and improving inference efficiency. dm for collab !
My vision is democratic AI where powerful models are not centralized, but accessible, understandable, and usable by all :D
| Project | Description | Demo / Repo |
|---|---|---|
| 🧪 PhysJEPA | Controlled JEPA vs pixel baseline diagnostic for naive physics — linear probes + VoE surprise in a 2D pymunk sandbox | Live Demo |
| 🍅 ActiveCrop | Uncertainty-aware active perception for safe VLM-based robotic crop intervention | Live Demo |
| ☀️ CausalSolar | Causal ML for weather impact on solar forecasting — interactive Playground + robustness suite | Live Demo |
| 🛡️ SafeAdapt | Research framework for studying alignment drift in continually interacting LLM agents | GitHub |
| 🧠 KAN-CNN | Interpretable image classification with Kolmogorov–Arnold Networks — edge functions, pruning & attribution | GitHub |
| 📄 OCR + OCRonos | RunPod serverless worker — PaddleOCR-VL-1.6 + OCRonos 4-bit text correction pipeline | GitHub |
| 🧾 Invoice Extractor | End-to-end invoice extraction — FastAPI, RapidOCR, fuzzy rules, Angular validation UI | GitHub |
| Project | Description | Demo / Repo |
|---|---|---|
| 🔗 Graph Agent | Config-driven LangGraph agent core — UI-free Python API + HTTP/SSE sidecar | GitHub |
| 📚 BookMind | RAG-powered companion for exploring great works of literature | GitHub |
| Project | Description | Deployed Demo |
|---|---|---|
| 🍃 Leaf | Interactive web app for geolocated map marker management | Leaf |
| 🧠 NN Visualiser | Visual, interactive explorer for neural network internals | NN Visualiser |
| 📦 Annotate Everything — BBox | Bounding-box annotation tool | BBox Tool |
| 🧊 Annotate Everything — Cube | 3D/cuboid annotation tool | Cube Tool |
| 📊 Annotate Everything — Table | Table structure annotation tool | Table Tool |
| 📚 Read & Track | Reading tracker and personal library management web app | Read & Track |
| 🏫 FST Simulation | 3D interactive simulation of the FST Beni Mellal campus — Babylon.js + Vue.js | GitHub |
| Model | Description | Deployed Demo |
|---|---|---|
| region-proposal-pp-doclayout-l-bo | Region proposal model for document layout analysis | 🚀 Demo |
| layoutlmv3-bo-segments | LayoutLMv3 fine-tuned for document segment classification | 🚀 Demo |
| SLANeXt_wired_bo_ft | Fine-tuned table structure recognition (wired tables) model | — |
| morocco-bo-ocr-latin-v3-finetuned | Fine-tuned OCR model on the French Moroccan Official Journal | 🚀 Demo |
| Invoice Layout Detection (YOLOv8n) | YOLOv8n model for invoice document layout detection | 🚀 Demo |
📚 Browse all my collections → — including my work on RL-MPQ LLMs (Reinforcement-Learning-based Mixed-Precision Quantization)
✨ Thanks for visiting hehe !