NeuroAI Lab Platform

A centralized data aggregation and analysis platform for neural timeseries data

Overview

This platform provides production-ready infrastructure for neuroscience research teams, showcasing end-to-end workflows from raw neural data ingestion to machine learning model training, with comprehensive experiment tracking. Built to mirror the internal tooling used by companies like Neuralink for managing large-scale neural recording datasets.

Key Features

Data Management

• Multi-Format Ingestion: Support for NWB, Parquet, JSON, and image formats
• Schema Validation: Automated validation for neural timeseries, metadata, and annotations
• Efficient Storage: Parquet-based storage with metadata indexing
• Quality Control: Real-time validation and quality metrics

Data Explorer

• Advanced Filtering: Search by subject, date, task, and custom metadata
• Feature Extraction: Bandpower analysis, spike detection, RMS amplitude
• Event Alignment: Trial-averaged responses aligned to experimental events
• Export Capabilities: Download processed data for external analysis

Visualization Dashboard

• Timeseries Viewer: Interactive multi-channel neural recording display
• Event Overlays: Experimental markers synchronized with neural data
• Spectral Analysis: Power spectral density across frequency bands
• Distribution Plots: Signal quality metrics and drift detection

Model Training

• Multiple Algorithms: Random Forest, SVM, XGBoost, PyTorch neural networks
• Cross-Validation: Robust k-fold CV with stratification
• Performance Metrics: Accuracy, precision, recall, F1, confusion matrices
• Model Versioning: Automated checkpoint saving and comparison

Experiment Tracking

• YAML Configurations: Reproducible experiment definitions
• Run History: Complete tracking of all training runs
• Metrics Storage: Structured metrics.json for each experiment
• Artifact Management: Organized storage of models, plots, and reports

Lab Registry

• Session Documentation: Rich notes with experimental observations
• Device Tracking: Recording device metadata and configurations
• Tagging System: Flexible categorization of sessions
• File Attachments: Protocols, analysis notes, supporting documents

Quick Start

Option 1: Docker Compose (Recommended for Recruiters)

Boot the entire platform with one command—no dependencies to install.

# Clone the repository
git clone <repository-url>
cd neuroai-lab-platform-demo

# Start all services (app, database, S3 emulator)
docker-compose up

The platform will be available at:

• Application: http://localhost:3000
• MinIO Console (S3 emulator): http://localhost:9001 (minioadmin/minioadmin)
• MySQL Database: localhost:3306 (neuroai_user/neuroai_password)

Helper Script (optional):

./docker-compose.sh start      # Start services
./docker-compose.sh stop       # Stop services
./docker-compose.sh logs app   # View app logs
./docker-compose.sh reset-db   # Reset database

Option 2: Local Development

Prerequisites

• Node.js 22+
• pnpm 10+
• MySQL/TiDB database

Installation

# Clone the repository
git clone <repository-url>
cd neuroai-lab-platform-demo

# Install dependencies
pnpm install

# Push database schema
pnpm db:push

# Start development server
pnpm dev

The platform will be available at http://localhost:3000

Production Build

# Build for production
pnpm build

# Start production server
pnpm start

Architecture

Frontend Stack

• React 19 with TypeScript for type-safe UI development
• Tailwind CSS 4 with custom neural-themed design system
• shadcn/ui component library for consistent UI patterns
• Recharts for interactive data visualization
• Wouter for lightweight client-side routing

Backend Stack

• Node.js + Express for server runtime
• tRPC for type-safe API layer with automatic client generation
• Drizzle ORM for database operations with MySQL/TiDB
• OAuth for authentication and user management
• S3-compatible storage for large dataset files

Design System

• Color Palette: Deep navy blue with electric blue accents
• Typography: IBM Plex Sans (UI) + IBM Plex Mono (code/data)
• Theme: Dark mode optimized for extended viewing
• Layout: Dashboard with persistent sidebar navigation

Docker Deployment

Services

The Docker Compose setup includes three services:

1. App (Node.js)

   • Runs the complete NeuroAI Lab Platform
   • Exposed on port 3000
   • Automatically builds from Dockerfile

2. MySQL Database

   • Stores all application data
   • Exposed on port 3306
   • Credentials: neuroai_user / neuroai_password
   • Database: neuroai_lab

3. MinIO S3 Emulator

   • Provides S3-compatible object storage
   • Web console on port 9001
   • Credentials: minioadmin / minioadmin
   • Useful for local development and testing

Project Structure

neuroai-lab-platform/
├── client/                 # Frontend application
│   ├── src/
│   │   ├── pages/         # Page components
│   │   │   ├── Home.tsx
│   │   │   ├── DataIngestion.tsx
│   │   │   ├── DataExplorer.tsx
│   │   │   ├── Visualization.tsx
│   │   │   ├── ModelTraining.tsx
│   │   │   ├── ExperimentTracking.tsx
│   │   │   └── LabRegistry.tsx
│   │   ├── components/    # Reusable UI components
│   │   ├── lib/          # Utilities and tRPC client
│   │   └── index.css     # Global styles and theme
│   └── public/           # Static assets
├── server/               # Backend API
│   ├── routers.ts       # tRPC procedure definitions
│   ├── db.ts            # Database query helpers
│   └── _core/           # Framework infrastructure
├── drizzle/             # Database schema and migrations
│   └── schema.ts        # Table definitions
├── shared/              # Shared types and constants
├── PROJECT_BRIEF.md     # Detailed feature documentation
└── README.md           # This file

Use Cases

For Research Labs

Provides a template for building centralized data management systems for neural recording experiments. Showcases best practices for reproducibility and experiment tracking.

For Neurotechnology Companies

Proves ability to translate complex scientific requirements into scalable, maintainable software infrastructure with professional UI/UX.

Development Workflow

Adding New Features

1. Update Database Schema (if needed)

   # Edit drizzle/schema.ts
   pnpm db:push

2. Create Backend Procedures

   // server/routers.ts
   export const appRouter = router({
     feature: router({
       list: protectedProcedure.query(async ({ ctx }) => {
         // Implementation
       }),
     }),
   });

3. Build Frontend UI

   // client/src/pages/Feature.tsx
   const { data } = trpc.feature.list.useQuery();

4. Write Tests

   pnpm test

Running Tests

# Run all tests
pnpm test

# Type checking
pnpm check

Demo Data

The platform includes mock data for demonstration:

• 3 recording sessions with multi-channel neural timeseries
• 2 trained models (Random Forest, SVM)
• 3 experiment runs with full metrics
• Sample configurations in YAML format

Design Philosophy

The platform follows a scientific/technical approach inspired by internal neuroscience tooling:

• Data-first: Emphasis on clarity and readability of neural data
• Minimal distractions: Clean interface that keeps focus on the data

Roadmap

Current Status (MVP)

• ✅ Complete UI for all core modules
• ✅ Data schemas and validation rules defined
• ✅ Visualization dashboards functional
• ✅ Experiment tracking interface
• ✅ Lab registry system
• ✅ Backend API implementation with tRPC
• ✅ Python data processing scripts (ingest.py, dataset.py)
• ✅ Docker containerization

Next Steps

• ML training pipeline (train.py, eval.py)
• CI/CD with GitHub Actions
• Demo video and screenshots

License

MIT License - see LICENSE file for details

Contributing

For production use, consider:

• Adding authentication middleware for multi-user support
• Implementing real data ingestion pipelines
• Integrating with actual neural recording hardware
• Adding more sophisticated ML models
• Implementing real-time data streaming

Contact

Built by Atin Srivastava.

---

Note: For production deployment with real neural data, additional security, validation, and compliance measures would be required.