BlockML – Visual Neural-Network Builder & Trainer

BlockML is a cross-platform desktop IDE that lets you design, validate, train, and package neural-network models through an intuitive, drag-and-drop interface.


✨ Core Highlights

  • Node-based editor
    Build computation graphs by connecting blocks that represent data loaders, layers, losses, optimizers, and more.

  • Real-time tensor validation
    The app continuously checks data-types, dimensional compatibility, and infers unknown shapes, surfacing errors or warnings inline.

  • Interactive parameter inspector
    Select any block to tweak hyper-parameters with type-safe controls (sliders, dropdowns, toggles).

  • Live training & analytics
    Start training inside the IDE and watch charts for loss, accuracy, and learning-rate schedules update in real time.

  • One-click export
    Package your model to portable formats such as ONNX for deployment across platforms and runtimes.


🛠️ Tech Stack

Layer / Concern Technology
Desktop shell Electron
Frontend React 18, Next.js 14, TypeScript
UI framework TailwindCSS with Radix UI primitives
Graph rendering ReactFlow
State management Zustand
Charting Recharts
Share this project:

Updates