Grant Lab Automation — AI Workflow Platform
An internal AI workflow platform with a visual builder for composing repeatable multi-step automations. It supports multiple model providers, real-time execution tracking, file processing, and administrative controls in one system.
Researchers needed a way to compose repeatable AI-assisted workflows without rebuilding prompts, file handling, execution tracking, and provider integrations each time.
Built the full-stack platform: workflow designer, provider abstraction layer, WebSocket execution tracking, file processing, admin tooling, and deployment.
Turned experimental AI-assisted tasks into a controllable internal platform instead of a scattered collection of one-off scripts.
Lead Developer
1 month
Reader-first case study
This case study focuses on problem framing, implementation choices, technical constraints, and outcome.
Lead Developer
1 month
Research project
2025
This is an internal workflow system. The case study focuses on platform design, provider abstraction, and execution control rather than private workflow definitions.
Visual workflow designer for multi-step AI runs
Supports Claude, OpenAI, and Gemini behind a unified execution layer
WebSocket execution tracking, file handling, admin controls, and audit logging
Project Overview
A full-stack AI workflow orchestration platform built to automate research tasks at the Grant Lab. Provides a visual interface for designing multi-step AI workflows, executing them in real-time, and managing results — all with enterprise-grade authentication and audit logging.
Project Details
Technologies Used
- Next.js 14 (frontend with Tailwind CSS)
- FastAPI (Python backend)
- PostgreSQL 15 (database)
- Zustand (state management)
- WebSocket (real-time execution updates)
- JWT + bcrypt (authentication)
- Docker Compose (deployment)
Key Features
- Visual Workflow Designer: Drag-and-drop interface for composing multi-step AI workflows
- Multi-AI Provider Support: Integrates Claude (Anthropic), OpenAI, and Google Gemini
- Real-Time Execution: WebSocket-based live progress tracking during workflow runs
- File Processing: Upload and process files through AI-powered pipelines
- Admin Panel: User management, API key configuration, system settings, and audit logs
- Security: JWT authentication with bcrypt password hashing
Technical Highlights
- Full-Stack Architecture: Decoupled Next.js frontend and FastAPI backend communicating via REST + WebSocket
- Multi-Provider Abstraction: Unified interface across Claude, OpenAI, and Gemini APIs
- Production-Ready: Dockerised deployment with PostgreSQL persistence and environment-based configuration