Weave - Drag & Drop Neural Networks
Hello everyone 
Just won the first place in Alexandria Cursor Hackathon as we have the best implementation using Convex.
We are introducing Weave - A modular drag-and-drop neural network editor that generates real PyTorch code. This is an MVP created during the hackathon to showcase how visual tools can accelerate ML experimentation.
The Problem:
Writing boilerplate PyTorch code for every experiment is repetitive and slow.
The Solution:
Weave lets you visually design neural network architectures, configure datasets, and train models - all from an intuitive drag-and-drop interface.
MVP Features:
• Drag-and-drop layer nodes (Linear, Conv2d, ReLU, Flatten, etc.)
• Real-time PyTorch code generation
• Live training with streaming logs and metrics
• Create reusable custom modules
• Checkpoint saving during training
• Dataset support (MNIST, CIFAR-10)
Tech Stack:
• Frontend: React + React Flow + Tailwind CSS
• Backend: Convex (TypeScript)
• Execution: FastAPI + PyTorch
Big thanks to Cursor Egypt Community for organizing this amazing hackathon!
Links:
• Demo: https://youtu.be/iH06YZBS38k?si=J0xI6IjKcZzetItX
• GitHub: https://github.com/Os14you/weave
Just won the first place in Alexandria Cursor Hackathon as we have the best implementation using Convex.
We are introducing Weave - A modular drag-and-drop neural network editor that generates real PyTorch code. This is an MVP created during the hackathon to showcase how visual tools can accelerate ML experimentation.
The Problem:
Writing boilerplate PyTorch code for every experiment is repetitive and slow.
The Solution:
Weave lets you visually design neural network architectures, configure datasets, and train models - all from an intuitive drag-and-drop interface.
MVP Features:
• Drag-and-drop layer nodes (Linear, Conv2d, ReLU, Flatten, etc.)
• Real-time PyTorch code generation
• Live training with streaming logs and metrics
• Create reusable custom modules
• Checkpoint saving during training
• Dataset support (MNIST, CIFAR-10)
Tech Stack:
• Frontend: React + React Flow + Tailwind CSS
• Backend: Convex (TypeScript)
• Execution: FastAPI + PyTorch
Big thanks to Cursor Egypt Community for organizing this amazing hackathon!
Links:• Demo: https://youtu.be/iH06YZBS38k?si=J0xI6IjKcZzetItX
• GitHub: https://github.com/Os14you/weave
Note: this was a vibe coding focus hackathon, just to expect some work arounds in the code and some weird bugs some times. but we are planning to fix and maintain it in the upcoming months.
