Cal Hacks转发了
AI Software Engineer Intern @ Fetch.ai | MS CS @ Santa Clara University | 13x Hackathon Winner | Specialized in Backend, Microservices, Distributed Systems, Java, Spring Boot, Python, MERN Stack, AWS, REST API
?? The Story of Autonomous AI Society ?? This weekend, I had the incredible opportunity to participate in Cal Hacks 11.0 ??, co-hosted by Fetch.ai and Google and organized by the University of California, Berkeley. I’m thrilled to share that my project, Autonomous AI Society, won the Fetch.ai Challenge. Despite my experience with numerous hackathons, finding a fresh idea felt challenging. Arriving late to the CalHacks Venue, I decided to embrace the opportunity to work solo and explore ideas without pressure. ?? I was exploring healthcare and education ideas but nothing clicked. Just as I was about to give up, news of Hurricane Milton in Florida and the rise of AI agentic technology sparked an interesting idea. With Fetch.ai providing a powerful framework for creating AI agents, I decided to simulate an AI society where AI agents were taking autonomous decisions during a disaster and trying to minimize the loss of life. It took 14 hours to build the proof of concept. I showed my first POC to Mark Losey and Sana Wajid CEng MICE FPWI who really liked the idea, that's when I decided to develop more features. How does it work? I developed 5 uAgents using the Fetch.ai framework, with each agent communicating with other agents through messages. The first agent initiates the entire disaster response process, next the distress analyzer agent takes different phone calls, performs distress analysis using Hume AI, identifies calls with highest distress, and decides to dispatch the drone to that caller's location. The next agent performs the drone simulation movement to the target location. Now the human detection agent gets the images from the drone and analyzes them using the Hyperbolic framework provided Llama vision model, to detect the humans and get the description of the image. Using Groq, the highest priority location is identified. Finally, the call rescue agent uses Llama 3.1405B model to make clear instructions for the rescue team and uses Vapi to make phone calls to the rescue team giving them directions and coordinates.?I used Deepgram for all voice functions. Demo of my project: Find in Comments???? Special thanks to Attila Bagoly and Chirag Maliwal for their continuous technical support during the hackathon. I am beyond grateful to Sana Wajid CEng MICE FPWI and Fetch.ai team for recognizing my project. It was a great learning experience working with different new AI technologies, thanks to the incredible sponsors Fetch.ai Fetch.ai Innovation Lab, Hume AI, Hyperbolic, Groq, Deepgram, Major League Hacking, Vapi. Listening to Perplexity CEO Aravind Srinivas and The Chainsmokers' Alex Pall share their journeys was a memorable highlight of the event. A special shoutout to the Cal Hacks team for bringing together the largest collegiate hackathon and creating such an amazing experience. #Hackathon #Innovation #CalHacks #Tech #AI #AIAgents #Coding #learning
-
+1