SoraChain DeSci Playground

Decentralized Science meets AI Collaboration

Explore how research data, training workflows, and AI models can be coordinated without central control – using Federated Learning + Blockchain.

Live Demo 👉 desci.sorachain.ai

⸻

🧠 What is the DeSci Playground?

The SoraChain DeSci Playground is an interactive sandbox to explore the future of AI research collaboration.

It shows how multiple researchers, institutions, or labs can: • Train AI models together (without sharing raw data) • Track contributions and model versions immutably • Earn rewards based on provable input or improvements

This prototype mimics a multi-stakeholder federated learning round, with real-time updates, node actions, and logs — all visible in one place.

⸻

🌍 Why It Matters

Today’s scientific research is: • Siloed across institutions 🏢 • Repetitive and redundant 🔁 • Centralized in terms of compute and recognition 🧠

SoraChain’s vision is to unlock collective AI training across silos, while guaranteeing: • Data privacy for all contributors • Provenance for training data and model deltas • Incentives for meaningful participation

This prototype demonstrates a future where AI models are public goods built through transparent and fair collaboration.

⸻

🔧 How It Works 1. Multiple nodes (simulated) train a shared AI model 2. Each node has its own local dataset 3. Training is done in rounds (Federated Averaging) 4. Contributions (weights, accuracy, etc.) are logged 5. Blockchain backend records events and signatures 6. Aggregator node merges and updates the global model

👉 You can view each node’s: • Dataset summary • Training logs • Accuracy changes • Blockchain events

⸻

🧪 Features • ✅ Interactive federated learning rounds • ✅ Per-node training logs • ✅ Visualized model accuracy trends • ✅ Blockchain-backed contribution tracking • 🚧 Coming soon: live stake delegation, model bounties, and researcher reputation system

⸻

🧱 Tech Stack • Frontend: React + TailwindCSS • Federated Learning Engine: PySyft + PyTorch • Blockchain Layer: Ethereum Testnet (custom contracts for provenance + rewards) • Backend API: FastAPI + WebSocket for updates • Model: Logistic Regression / Simple Classifier for demo

⸻

🔭 Use Cases We’re Exploring • Collaborative training on rare disease datasets • Model development across climate sensor networks • Secure AI in medical research (FL + privacy-preserving ML) • Open-source scientific benchmarks + leaderboards

⸻

👩‍🔬 For Researchers - Join us for Pilot

Want to plug in your dataset? Join a federated round? We’re building tooling to let you join with one click — while keeping your data local.

Email us at: swayam@sorachain.ai