Verda Compute Track

INTRODUCTION

From electron to token, it takes decades of progress to run AI. Add scale, efficiency, and sustainability, and AI becomes a monumental achievement.

Verda brings everything together.

We make compute, storage, and networking work in unison to squeeze peak performance, hosted in European data centers and utilizing 100% renewable energy. We develop services for you to utilize computational power most efficiently, no matter your use case.

The end result is a cloud platform trusted by the world's leading AI labs and frontier researchers. The best thing is you get to use the same next-gen compute for free at Junction 2026.

BOUNTY: BEST USE OF VERDA

Pick any challenge at Junction and use Verda products to build something you couldn't build on a laptop. Then show us how far it goes.

• Teach a humanoid to move. Train locomotion or manipulation policies for Aalto's Unitree G1 in simulation, at scale, before your shot at the real robot.
• Turn demonstrations into a brain. Train imitation-learning or VLA policies from robot-arm demos in the Ignite × Intel challenge, or from Rokoko motion-capture teleoperation data.
• Make an agent stronger than it was on Friday. Run thousands of rollouts in your AfterQuery RL environment and fine-tune a smaller model on the experience.
• Read the whole internet's encyclopedia. Embed MariaDB's full Wikipedia dataset with batch jobs and serve a RAG assistant on top.
• Keep sensitive data in Europe. Run open-weight models for a banking coach or a recruitment tool on infrastructure you control, instead of sending data to someone else's API.

Or surprise us. Anything that needs serious compute counts. (Anything except crypto mining. We've seen it, and it gets you banned.)

What we're looking for

• An ambitious idea where compute is what makes it possible
• The right Verda product for the job, and a clear reason why you chose it
• Your path from first idea to working prototype
• Bonus points: show how your solution would scale from a hackathon demo to real users

Minimum Requirements

• Document your process
• Share a link to a summary → Post it in our Discord channel

The Technology

• Inference API for serving out-of-the-box open-weight models within your application.
• GPU instances or Instant clusters for training or fine-tuning tasks.
• CPU instances for hosting agentic applications.
• Auto-scaling containers for hosting models where scale and costs matter.
• Batch jobs for long-running tasks in the background.

Prizes and Judging

The winning team gets 10,000 USD in cloud credits and Verda hoodies. Runner-up teams in second and third places get 2,500 USD in cloud credits.

What we'll bring

• Inference API: serve open-weight models in your app with zero setup
• GPU instances and Instant Clusters: for training and fine-tuning when one GPU isn't enough
• CPU instances: to host your agents and backends
• Auto-scaling containers: deploy models where scale and cost matter
• Batch jobs: run long jobs in the background while you sleep (or don't)
• NVIDIA Nemotron endpoint: hosted by Verda together with NVIDIA
• Free credits for every team: scan the QR code on your table or go to verda.com/blog/junction-2026

Mentors on-site during the entire weekend, meet us on the third floor at Booth 6! Start with the "Getting started with Verda" workshop on Friday at 20:30, right after the opening ceremony: sign-up, credits, every product, and how to use them.

Ambition: how much did compute let you do that you otherwise couldn't? Fit: did you pick the right Verda services and use them well? Working prototype: does it run, and is the process documented? Path to scale: is there a credible way from demo to real users?