🚀 Australia's AI Research Hackathon
Join us for a weekend solving some of the most interesting and cutting edge problems in AI, from RL and agents to mechanistic interpretability.
What We're Building
Whether you're an engineer who wants to work on improving AI safety, a student building your first agent, or someone who's been itching to train your own model for a novel use case, we want you to come and get stuck in!
With generous compute credits for each participant courtesy of our friends at Runpod, you can actually train real models over the weekend — with the winning team taking home $2,500 worth of compute credits!
We also have an awesome set of judges lined up for this event including AI researchers from Canva, Exa Labs and EleutherAI who will be around over the weekend to help answer questions and provide guidance on project ideas.
Don’t have a problem in mind to solve yet? No worries, we have a few categories you can choose from:
🔬 Mechanistic Interpretability
Pro-human AI, AI safety & mechanistic interpretability of neural networks. Understand what’s really happening in neural networks and help build systems that support human flourishing!
Example projects: Check out Lucia’s project ideas here
New to interp? Check out: Anthropic's transformer circuits | Neel Nanda's tutorials
🤖 AI Agents in the Wild
AI agents have shown remarkable potential at solving real world problems either autonomously or assisting a human, particularly in the domains of coding, customer service and personal assistance. We believe there are many more domains AI agents have yet to transform.
Your mission: This category encourages you to pick an area you’re interested in and build an agent to automate it or assist humans with it.
🧠 Long-Horizon RL Training
Models like DeepSeek R1 and OpenAI o1/o3 proved that long-horizon RL (RLVR) can drastically improve performance in a specific domain such as coding or maths.
However, creating high quality datasets outside of math and code remain an unsolved problem, and for long running tasks, creating intermediate RL rewards is also an area of active work.
Your challenge: This category encourages you to pick a task you’re interested in training models on, and come up with a smart new way of automatically generating the data needed for RL training.
The Details
📍 Venue
An awesome venture studio and co-working space in the heart of Redfern.
Bring your laptop and curiosity, we'll provide food, caffeine, and good vibes!
💻 Compute Credits
Thanks to our wonderful friends at RunPod, every participant will get $50 in compute credits to use across all Runpod services over the weekend!
👥 Teams
Come solo or with friends. Need a team? We'll sort it out during kickoff!
🏆 Prizes:
Grand Prize: $2,500 of Runpod credits
+ Additional prizes TBA (watch this space!)
Judges
Ethan Smith — Lead AI Researcher @ Canva
Ethan co-founded Leonardo.Ai, where he was the head of research.
William Li — Member of Technical Staff @ Exa Labs
Will was a senior ML engineer at Atlassian and led development of Rovo Deep Research.
Lucia Quirke — Member of Technical Staff @ EleutherAI
Lucia was an engineer at Atlassian and Canva before transitioning to AI research through the MATS scholarship.
Who Should Join?
Engineers & researchers interested in AI
Students eager to ship real projects
Designers who want to shape AI experiences
Domain experts looking to apply AI to their field
Anyone excited about pushing the frontier!
Schedule
Saturday, November 22nd
8:00 AM — Doors open | Coffee, chats & intros
9:00 AM — Hackathon kickoff | Team formation
12:00 PM — Lunch
6:00 PM — Dinner
9:00 PM — Doors close (or take it home!)
Sunday, November 23rd
8:00 AM — Doors open
1:00 PM — Lunch
4:00 PM — Submissions due
5:00 PM — Team pitches
6:00 PM — Judging & prizes
Sponsors
Huge thanks to RunPod for providing compute credits and prize money! 🎉
And to the awesome team at Helix Collective for letting us use their incredible space!
🎯 Interested in sponsoring? Reach out to [email protected]
Questions?
DM us!