Google Physical Engineering AI Agent Hackathon
📌 About the Event
Google – Agent Harness on Physical-Engineering’s Last Exam is a global AI hackathon organised by Google in collaboration with Mercor. Participants are challenged to build autonomous AI agents capable of solving complex engineering tasks across mechanical, electrical, civil and structural engineering. The competition evaluates AI agents on their ability to analyse technical documents, perform calculations and generate practical engineering solutions. Participants can compete for a share of the $400,000 prize pool.
ℹ️ Event Details
- Organizer: Google (in collaboration with Mercor)
- Mode of Conduct: Online
- Venue: Online
- Competition Type: Global AI Hackathon
- Team Size: Up to 5 members
- Registration Fee: Free
📅 Important Dates
- Competition Announcement: 3 November 2026
- Registration/Competition Start: 9 November 2026
- Final Submission Deadline: 6 January 2027, 5:29 AM IST
- Stage 1 Writeup Judging Complete: 19 January, 2027
- Stage 2 Test Set Evaluation Complete & Winners Announced: 26 January, 2027
🎯 Eligibility & Rules
- Suitable for AI developers, machine learning enthusiasts, software developers and engineering students.
- Participants must develop an autonomous agent harness using one of the Google-sponsored models.
- Submissions must follow the prescribed competition protocol and technical requirements.
- Participants must submit their working agent harness, development-set evaluation artifacts and a Kaggle write-up.
- The solution must be tested using the provided evaluation kit before submission.
🏆 Benefits for the Participants
- Total Prize Pool: $400,000 USD.
- 1st Prize: $250,000 USD
- 2nd Prize: $100,000 USD
- 3rd Prize: $50,000 USD
- Opportunity to compete in a global AI hackathon organised by Google.
- Opportunity to develop practical skills in AI agents, machine learning, automation and engineering problem-solving.
- Hands-on experience working with engineering datasets, CAD/BIM-related files, technical documents and computational tools.