Automated Data Scientist (GeneEase)

Inspiration

• Recent advances in AI using multiple agents and orchestrating tasks with a team of agents
• The revolutionary potential of applying AI-driven data science to revolutionize what we know about the human genome

What it does

• Multiple agents orchestrate with one another to plan a project, develop and debug code, interpret results - presented through both a chatbot interface and a mind mapping interface, allowing users to explore genomic data sets in entirely new ways

How we built it

• The engine behind GeneEase orchestrates multiple agents in succession to move from input query to insights with visualizations. The back end involves agents discussing with one another to think through what variables are important, develop code, test and run code, then interpret output results. The front end mind map was built with different nodes that communicate with our multi-agent system to develop and visualize multiple analyses.

Challenges we ran into

• Getting the agents to orchestrate properly
• Integrating the mind map UI

Accomplishments that we're proud of

• Bringing the project together to have the back end agents collaborating seamlessly in time
• Developing both a chat-based interface and a mind mapping interface

What we learned

• Multi-agent orchestration has the potential to revolutionize the future of data science and analysis, democratizing the ability to sort through large datasets
• Our process for exploring large data sets is non-linear and very different from how current tools are designed

What's next for Automated Data Scientist (GeneEase)

• Further expanding the range of acceptable data from genotype to full genomic
• Adding more collaboration features
• Including more agents in the multi-agent orchestration for automated discovery of insights

Built With

• amazon-web-services
• autogen
• javascript
• openai
• python
• superbase
• typescript