CAMEL’s Mastering Multi-Agent Systems Hackathon, held in Oxford, brought together tech enthusiasts, academics, and AI professionals to explore and create innovative multi-agent solutions. The event combined workshops, networking, and a hackathon that ignited creative approaches to multi-agent applications.
Hackathon Highlights
1. Introduction to CAMEL Multi-Agent System
Our founder, Guohao Li, opened with CAMEL’s mission: Finding the scaling laws of agents.
Prof. Philip Torr (Professor at the University of Oxford and Chief Scientific Advisor at CAMEL AI and Eigent AI) delivered an inspiring talk on the future of agentic AI.
2. Hands-On Learning with CAMEL Workshops
Workshops led by CAMEL experts Roman, Sally and Caelum gave participants hands-on experience with the CAMEL framework on Google Colab, empowering them to design and develop multi-agent models.
-min.jpeg)
A Look at Each Track and Winning Projects
The hackathon featured four tracks, each supported by key partners and offering substantial prizes. Here’s a breakdown of the winners in each track:
⚡️ "Lightning Quick Agents" Track – Sponsored by SambaNova AI
1st Place: JazzCoders by Ben Hardin and team
A real-time jazz improvisation model where agents “listen” to each other and create spontaneous solos, capturing the spirit of live jazz.
2nd Place: AutoWeatherForecaster
A tool that uses multi-agent systems to predict weather patterns based on real-time data.
3rd Place: Best Stock Pricing System of All Time & It Is Compliant by K2024HU, Caelum Forder, and Qianshuo Ye
A stock-pricing app that assesses if a stock price is fair, too high, or too low.
📄 "Local Document Mastery" Track – Supported by Qdrant Engine & Chunkr.ai
1st Place: Podstyle by Moncef Slimani
An AI that mimics celebrity dialogue, allowing users to engage in interactive, personalized conversations.
2nd Place: Music KG by Pauline Sepu and team
An AI system that creates a knowledge graph for music genres and artists, offering users a unique music discovery experience.
3rd Place: Patient-Doctor Multi-Agent System by Zijian Ling and team
A multi-agent simulation for patient-doctor interactions across various demographics and specialities.
🏆 "Agents with Frontier Models Innovation" Track – Partnered with Mistral AI
1st Place: Symposion: Your Research Proposal Companion Tool by Samuele Marro, J Rosser, and Jarek Liesen
An AI that accelerates research proposal creation, offering simulated expert feedback, collaboration modeling, and funding suggestions.
2nd Place: Auto Paper Review Bot by Bohui Zhang, Yu Liu, and Kaiwen Zuo
A bot that reviews academic papers, providing expert feedback on key sections like abstract, methodology, and conclusions.
3rd Place: Automated Traditional Chinese Medicine Diagnosis by Chen Ziji, Zhaocheng Gong, Yu Han, and Zexi Li
An AI that processes patient data to generate TCM diagnoses, incorporating factors like symptoms, prescriptions, pulse, and tongue conditions.
🔥 "Connecting to the Web" Track – Sponsored by FireCrawl
1st Place: GrantMe by Tim Tan, Thierry Blankenstein, Jaron Cohen, and Nicole Kagan
A multi-agent system that manages grant applications, automating funding searches and organizing data to streamline the research process.
2nd Place: Coconut - Movehouse.ai by Charlie Cheesma, Raymond, Tamzin, and George
Movehouse.ai simplifies house-hunting by using multi-agent systems to find, assess, and book properties that meet users’ needs.
3rd Place: The Bid Whisperer by Celine Xie and Douglas Lai
This tool addresses the “winner’s curse” in auction markets by using agents to gather and analyze data, providing more accurate market valuations for collectable items.
For more information, please check out our Devpost page to dive into each hackathon creation!
https://mastering-multi-agent-system.devpost.com/project-gallery
Each project showcased the power of multi-agent systems to solve practical challenges creatively and efficiently. From music generation to research facilitation, the hackathon emphasized CAMEL’s mission to find scaling laws of agents.
Special Thanks
This event couldn’t have been possible without our partners and sponsors: SambaNova AI, Mistral AI, FireCrawl, Qdrant Engine, Chunkr.ai, Oxford AI society, Oxford Founder Society and Oxford CSSA. Their support helped make each track a specialized domain, allowing participants to explore how multi-agent systems can make real-world impacts across diverse fields.
