Imperial AI Hackathon is Imperial AI Society's flagship hackathon and the finale of Imperial AI Week. Over one weekend, 450+ builders in teams of up to 4 take on five AI tracks, with Amazon voucher prizes and partner bounties.
Saturday 14 - Sunday 15 November 2026
Huxley Building, Imperial College London, South Kensington
Pick one track. Every track pays £500 / £200 Amazon vouchers for 1st / 2nd, and you can go for any bounty inside it.
Autonomous Agents (Headline Track)Agents that research, write and run code, and reliably finish real multi-step tasks.
Track winners - £500 / £200
Databricks bounty - £960 + swag. Build on the Databricks Data Intelligence Platform: a pipeline, agent or app. Scored on creativity, technical depth and real-world impact.
Best use of Linkup - £560 + credits. An agent built around Linkup live web search that researches, verifies or acts on fresh information, citing its sources.
Best use of Synthesia - 50,000 credits. The best product built with Synthesia's Interactive Avatar API.
Best use of Tavily - 30K / 20K / 10K credits. Best use of the Tavily search API, for example a stale-answer detector or a pre-meeting briefing agent.
MongoDB - Atlas voucher. An AI app or agent with MongoDB Atlas at its core, such as Vector Search or agent memory.
Voice, Vision & Generative Media (Gold Track)AI products people talk to, watch or create with: voice, avatars, images and video.
Track winners - £500 / £200
Best ElevenLabs project - Scale tier. Real-time voice agents beyond plain text-to-speech. Scored on agent depth, interaction design, technical integration and novelty.
Best use of fal - $1,000 + merch. Best use of fal's generative media models: image, video or audio.
Best use of Anam - 1-yr Growth plan. Avatar agents built with Anam using skills and tool calling. Rewards creativity, realism and multi-modal interaction.
AI for Science (Gold Track)Use AI to speed up how we understand and design molecules, cells and materials.
Track winners - £500 / £200
Best use of Boltz - Boltz credits. Best use of Boltz's biomolecular models: structure, binding affinity or molecule design.
AI Poker Bots (Gold Track with AlgoSoc)Build an AI agent that plays no-limit Texas Hold'em and beats the other teams' bots in a live tournament. No judges: the tournament decides.
Track winners - £500 / £200
1st place extra - 4 Optiver speakers. Goes to the winning bot.
Reply Challenge: AI on Campus (Gold Track)The challenge. Design an AI-based solution for a real university problem, built as a working prototype with present state, guardrails, human review and real-time interaction, via no-code orchestration or full-stack. Advanced teams can deploy a multi-agent system with specialised agent roles, dynamic mode-switching and human-in-the-loop escalation. Judged by Reply.
1st place - £500 + office visit. £500 Amazon vouchers, 4 backpacks and an office visit to Reply Labs in London.
2nd place - £200. £200 Amazon vouchers and 4 water bottles.
Overall prizes (open to every team, any track)Best-Designed AI Product (Figma) - £960 + merch. An AI product as well designed as it is smart, designed in Figma or built with Figma Make.
Best use of Adaption - $1,000 credits each + merch. Fine-tune a model with Adaption. Ranked on Adaption's leaderboard by win rate against the base model.
Partners & SponsorsHeadline: Adaption
Gold: Anthropic, Figma
Silver: Optiver
Bronze: Jane Street, tbd.fund
Bounty: Databricks, Linkup
Supporting: Vita Coco
Tech: fal, ElevenLabs, Tavily, Lovable, Synthesia, Anam, Boltz, MongoDB, Tamarind Bio
Community: Entrepreneurs First, EWOR, Google Developer Groups
Schedule Saturday 14 November
08:00 Partner setup
09:00 Check-in opens
10:00-10:30 Welcome
10:30-11:00 Adaption online workshop
11:00-11:05 Figma Best-Designed prize
11:05-11:25 Reply Student Tech Clash
11:25-11:40 Autonomous Agents track and partner briefs
11:40-11:50 Voice, Vision and Generative Media track and partner briefs
11:50-12:00 AI for Science track
12:00-12:10 AI Poker Bots
12:10-13:00 Team formation, building starts
13:00 Lunch
15:00-15:30 Reply workshop
15:30-15:55 Linkup talk
19:00 Dinner
21:00 Building closes
10:00 Doors open, breakfast
12:00 Submissions close
12:30 Judges arrive, 12:45 briefing
13:00-16:30 Judging: we play your 2-minute video, then the judges question your team
15:50-16:30 Live finals for the top Autonomous Agents and Voice teams
17:00 Finals on the main stage
17:30 Prizes, finishing around 18:00
£ prizes are Amazon vouchers. Submissions close Sun 12:00 · judging 13:00 · finals 17:00.
QuestionsEmail [email protected]
Requirements
Every team submits ONE project on Devpost by 12:00 on Sunday 15 November. Late submissions are not judged, and nothing can be changed after 12:00.
Your submission must include:
- A demo video of up to 2 minutes. This is what the judges watch in your slot, so show the product working. You can edit it and speed parts up, but label any sped-up section (for example "4x") and say how long a real run takes. Videos are cut off at 2:00. No video means the project is not judged.
- A public GitHub repository with your code. Your last commit must be before 12:00.
- Your track (pick one) and the bounties you are entering. You can enter up to 2 bounties from your own track. Figma and Adaption are open to every track and do not count towards the 2.
- Every team member added to the Devpost project, with each person's full name, Luma signup email and university email in the team members question.
- If you are entering Best use of Adaption: your Adaption model ID, exactly as shown on the Adaption leaderboard. If you are entering Best-Designed AI Product: your Figma file or Figma Make link.
What to show in your 2 minutes. The judges mark what they see in the video and what you say afterwards, so use the time like this:
- First 15 seconds: the problem. Say who it is for and what problem it solves. Name a specific user, not "everyone".
- About 75 seconds: the product doing the whole task. Show one real run from start to finish on screen. This carries the most marks. Say what is automated and what is manual or mocked.
- About 20 seconds: how it is built. Name the main parts and why you chose them. Show on screen where each bounty tool you entered is used.
- Last 10 seconds: what is new. Name the closest existing product and how yours differs, and say how often your product works when you run it.
Screen recordings with a voice-over work best. Skip logos, intros and background music.
AI Poker Bots teams also upload their bot to the AlgoSoc tournament system by 12:00.
In your judging slot: we play your video, then the judges ask your team questions for 6 minutes. There are no slides and no live demo in this round. Bring a laptop with the output of a finished run open, in case a judge asks to see it.
Prizes
Autonomous Agents - 1st place: £500 Amazon vouchers
£500 Amazon vouchers. Track: Build agents that reliably complete multi-step tasks.
• Databricks - £960 + swag
• Linkup - £560 + credits
• Synthesia - 50,000 credits
• Tavily - 30K / 20K / 10K credits
• MongoDB Atlas - voucher
Autonomous Agents - 2nd place: £200 Amazon vouchers
£200 Amazon vouchers. Track: Build agents that reliably complete multi-step tasks.
• Databricks - £960 + swag
• Linkup - £560 + credits
• Synthesia - 50,000 credits
• Tavily - 30K / 20K / 10K credits
• MongoDB Atlas - voucher
AI for Science - 1st place: £500 Amazon vouchers
£500 Amazon vouchers. Track: Use AI to understand and design molecules, cells and materials.
• Boltz - credits
AI for Science - 2nd place: £200 Amazon vouchers
£200 Amazon vouchers. Track: Use AI to understand and design molecules, cells and materials.
• Boltz - credits
Voice, Vision & Generative Media - 1st place: £500 Amazon vouchers
£500 Amazon vouchers. Track: Build AI products people talk to, watch or create with.
• ElevenLabs - Scale tier
• fal - $1,000 + merch
• Anam - 1-year Growth plan
Voice, Vision & Generative Media - 2nd place: £200 Amazon vouchers
£200 Amazon vouchers. Track: Build AI products people talk to, watch or create with.
• ElevenLabs - Scale tier
• fal - $1,000 + merch
• Anam - 1-year Growth plan
AI Poker Bots - 1st place: £500 Amazon vouchers + 4 Optiver speakers
£500 Amazon vouchers.
Track: Build an AI bot to compete in a live no-limit Texas Hold’em tournament.
• First-place extra - 4 Optiver speakers
AI Poker Bots - 2nd place: £200 Amazon vouchers
£200 Amazon vouchers.
Track: Build an AI bot to compete in a live no-limit Texas Hold’em tournament.
Reply Challenge: AI on Campus - 1st place: £500 Amazon vouchers + office visit
£500 Amazon vouchers. Track: Build an AI solution for a real university problem.
• 1st place - £500 vouchers, 4 backpacks + Reply Labs London office visit
• 2nd place - £200 vouchers + 4 water bottles
Reply Challenge: AI on Campus - 2nd place: £200 Amazon vouchers
£200 Amazon vouchers. Track: Build an AI solution for a real university problem.
• 1st place - £500 vouchers, 4 backpacks + Reply Labs London office visit
• 2nd place - £200 vouchers + 4 water bottles
Best-Designed AI Product (Figma) - £960 + merch
Open to all teams across every track.
• Best-Designed AI Product (Figma) - £960 + merch
Best use of Adaption - $1,000 credits each + merch
Open to all teams across every track.
• Best use of Adaption - $1,000 credits per team + merch
Judges
To be added
Judging Criteria
-
Track mark scheme
Each criterion is marked 0 to 3 and weighted by track. The criteria and weights for each track are in the Rules. Bounties are judged on how central the partner's tool is to your product. -
What to show in your 2-minute video
Show one real run of the whole task from start to finish. Say who it is for, what is automated and what is manual, and how often it works. Show where each bounty tool is used. Label sped-up parts.
Questions? Email the hackathon manager