Logo

LabTeam: Multi-Agent Lab Simulation for Education & Research

🧠 Inspiration

Many schools and research institutes lack access to expensive lab automation tools or AI-powered workflows. Inspired by the potential of generative AI + 5G/Edge computing, we wanted to create an open-source, multi-agent lab simulation environment that allows students and researchers to explore scientific workflows in an intuitive, hands-on, and cost-effective way.

⚙️ What it does

LabTeam simulates a real-world laboratory with multiple collaborating AI agents:

Atlas: Project manager orchestrating all agents
Bohr: Literature review and hypothesis generation
Curie: Bayesian optimizer suggesting experiment parameters
Deng: Generates robotic lab protocols
Edison: 3D modeling of labware components

Users interact with the system via a clean web interface that integrates with the ADK Agent UI via iframe. Execution status, experiment logs, and results are presented visually in real time.

🏗️ How we built it

Frontend: React + Tailwind CSS + Framer Motion for dynamic UI
Backend: Google ADK (Agent Development Kit) running AI agents
LLM Integration: Used Amazon Bedrock to power Claude 3.5 agents via LiteLLM wrapper
Inter-agent communication: Handled via ADK sub-agent orchestration
Hosting: Local development with edge/5G compatibility for future deployment

🧱 Challenges we ran into

Integrating Amazon Bedrock authentication and finding correct model access permissions
Mapping agent instructions and outputs to intuitive UI elements
Ensuring iframe-based UI communication remained reliable across local ports
Balancing modular agent design with cross-agent coordination logic

🏆 Accomplishments that we're proud of

Fully integrated Claude 3.5 on Amazon Bedrock with ADK agents
Developed a reusable LLM wrapper to swap backend models easily
Built an interactive pipeline dashboard with real-time status tracking
Delivered a project that could scale across classrooms, labs, and low-cost hardware