Helixai

Helixai is a Decentralized Science (DeSci) platform that democratizes and brings transparency to scientific research.

Core Problems We Solve: Traditional scientific research relies on centralized institutions, facing issues with data reproducibility and transparency. Additionally, general researchers struggle to access expensive analytical tools.

Our Solution:

1. AI-Powered Research Tools: By integrating Stanford's Biomni AI Agent, anyone can perform gene analysis, drug repurposing, and biomedical research using natural language queries.

2. Permanent Blockchain Storage: All research results are stored on the Base Sepolia blockchain, creating verifiable and tamper-proof scientific records.

3. Real-time AI Interaction: Through Server-Sent Events (SSE), users can engage in real-time conversations with AI while conducting research, transparently viewing the reasoning process.

4. Open Ecosystem: Anyone can participate by simply connecting their wallet, and all research results are verifiable on the Base Sepolia block explorer.

Use Cases:

• Analyzing the association between BRCA1 gene and breast cancer
• Discovering novel drug targets for Alzheimer's treatment
• Studying protein folding patterns in rare genetic disorders

Vision: We aim to transform scientific research from an activity controlled by a few institutions into an open ecosystem where anyone worldwide can participate and contribute, ultimately advancing human health and welfare.

Architecture Overview: Helixai is built on a 3-tier architecture: Frontend (Next.js), Backend (FastAPI + Biomni AI), and Blockchain (Base Sepolia).

Frontend (Next.js 15.2.4):

• Scaffold-ETH 2 Framework: Provides cutting-edge tools for Ethereum dApp development. Implemented type-safe Web3 interactions using Wagmi v2 and Viem.
• React 19 + TypeScript: Optimized performance using the latest React Server Components and App Router.
• RainbowKit: Simplified multi-chain wallet connection implementation.
• shadcn/ui + Tailwind CSS: Delivers consistent and beautiful UI/UX.
• Recharts: Implemented interactive charts visualizing research activity data.

Backend (Python FastAPI):

• Biomni AI Agent Integration: Uses Stanford's biomedical research-specialized AI agent as the core. This agent can automatically utilize over 40 bioinformatics tools.
• Server-Sent Events (SSE): Implemented real-time streaming responses to transparently show users the AI's reasoning process. This is a key feature enhancing scientific research reliability.
• Session Management: Maintains each user's conversation context to enable continuous research queries.
• Web3.py: Implemented a service that automatically stores research results on the blockchain.

Blockchain (Base Sepolia + Hardhat 3):

• Why Base Sepolia: An Optimism-based L2 network providing low gas fees and fast transactions, with stability guaranteed by Coinbase support.
• Hardhat 3 Upgrade:
  • 3x faster builds with parallel compilation (12.3s → 4.1s)
  • 2x faster recompilation with incremental compilation
  • Improved developer experience with native TypeScript support
  • 38% reduced memory usage (450MB → 280MB)
• ResearchRegistry Smart Contract: Permanently stores research title, description, IPFS hash, researcher address, and timestamp.

Notable Implementation Highlights (Hacky but Notable):

1. SSE-based Real-time AI Streaming: We combined FastAPI's StreamingResponse with Next.js's ReadableStream to deliver the AI's reasoning process to users in real-time. We manage states granularly with various event types like event: message, event: blockchain, event: done. This provides a much richer user experience than typical REST APIs.

2. Biomni AI Multi-stage Reasoning Visualization: We show users everything about what tools the AI uses internally and what thought processes it goes through. Implemented with Collapsible UI so detailed logs can be checked only when needed. This is a key feature ensuring reproducibility and transparency in scientific research.

3. Automated Blockchain Storage Pipeline: When AI research completes, the backend automatically stores results on the blockchain via Web3.py. Users can permanently preserve results without additional wallet signatures. We display transaction hash, block number, and gas usage in real-time on the frontend.

4. Scaffold-ETH 2 Custom Hooks Utilization: We implemented type-safe and automated contract interactions using hooks like useScaffoldReadContract and useScaffoldWriteContract. ABI and addresses automatically synchronize, significantly improving frontend development speed.

Technology Stack Summary:

• Frontend: Next.js 15.2.4, React 19, TypeScript, Wagmi v2, Viem, RainbowKit, shadcn/ui
• Backend: FastAPI, Biomni AI Agent, Web3.py, Server-Sent Events
• Blockchain: Solidity 0.8.20, Hardhat 3, Base Sepolia (Optimism L2)
• AI Models: Claude 3.5, GPT-4o, Gemini Pro (Biomni supported)
• Infrastructure: Yarn Workspaces monorepo, Vercel (frontend), Self-hosted (backend)

This project is not just a combination of technologies, but a platform that organically integrates three innovative fields - AI, blockchain, and bioinformatics - to present the future of scientific research.