[ online hackathon ] / deadline: 20 september 2026, 23:59 cet

Build something we couldn’t have predicted.
We’ll handle the infra.

Free credits across 11 clouds and 23 GPU models for the duration + $50,000 in prizes.

Apply for hackathon What to build ↓
emma agent CLI
$ emma run "spin a 4-core 16GB spot in Europe under $0.05/hr"
→ scanning 11 providers · 4 regions · 23 GPU models
region provider shape price
fr-par-1lambda·labs1×H100$2.39/hr
de-fra-2runpod1×H100$2.49/hr
nl-ams-1voltage·park1×H100$2.55/hr
us-east-2coreweave1×H200$3.79/hr
fi-hel-1datacrunch8×B300$28.40/hr
✓ picked fr-par-1 · lambda·labs · $2.39/hr
→ provisioning…
✓ ssh root@198.51.100.42
cold→ssh · 38s 11 providers · live

This is emma. Multi-cloud GPU infra you can talk to.

01 / tracks

Two tracks. Pick one.

track 01

2–3 weeks 1–5 people

Builders

Long, team-friendly — provisioning to maintenance.

You ship a working product on emma’s API/SDK. Real provisioning, real monitoring, real maintenance. Multi-cloud is the deliverable, not a checkbox.

Examples, not requirements · 4 ideas

  • // build E2B on emma

    Multi-cloud code execution sandbox — build E2B on emma. AI agents need a safe place to run generated code. POST a sandbox request, get a WebSocket URL back. Pre-warmed VMs across the cheapest providers. Aim for 3–5× cheaper than E2B.

  • // build Lambda on emma

    Multi-cloud FaaS — build Lambda on emma. Functions deploy via upload, scale via warm pool, bill per invocation. Multi-region routing picks the cheapest active cluster. Compare your cost graph to Lambda’s, on-screen.

  • // build Deadline Cloud on emma

    Distributed render farm on spot — build Deadline Cloud on emma. Blender scene goes in. Spot VMs spin up across 4–5 providers simultaneously. Frames distribute, render, stitch. 60–80% cheaper than AWS Batch because spot spans the whole emma market.

  • // build CAST AI on emma

    Autonomous cloud optimizer — build CAST AI on emma. An agent that monitors all your resources, proposes optimizations through a Slack approval flow, executes them, and reports weekly savings. Cross-provider, not single-cloud.
Judged on · 4 criteria

  1. 01

    Does it run in production?

    Working product on emma’s API/SDK. Real provisioning, real logs.

    Required: GitHub repo + working product + demo video

  2. 02

    Is emma essential?

    Couldn’t have been built on a single cloud.

  3. 03

    Is it maintainable?

    Recovery, monitoring, error handling.

  4. 04

    Would you ship it as a product?

    Could survive a Monday standup with engineering.

track 02

a long weekend 1–2 people

Showcases

Short, solo-friendly, video-first.

You demo something that goes from prompt to running result in under 2 minutes. Clean. Visual. Shareable. Record it once, post it, the world sees what emma does.

Examples, not requirements · 4 ideas

  • $ “Give me a 4-core 16GB spot in Europe under $0.05/hr.”

    Live spot price auctioneer. Prompt: ‘Give me a 4-core 16GB spot in Europe under $0.05/hr.’ Live-rendered table of 10+ providers ranked by current price. One keypress provisions. Under 90 seconds from prompt to provisioned VM.

  • $ “Spin me up a Python dev box in Germany with at least 8GB RAM.”

    Natural language VM provisioner. Prompt: ‘Spin me up a Python dev box in Germany with at least 8GB RAM.’ Sentence in. SSH command out. Under 2 minutes. Optional cloud-init bonus.

  • $ “Deploy ghcr.io/user/myapp:latest to Kubernetes, cheapest option.”

    One-command app deploy to Kubernetes. Prompt: ‘Deploy ghcr.io/user/myapp:latest to Kubernetes, cheapest option.’ Provisions K8s on the cheapest active region, applies your manifest, returns a live URL.

  • $ “Show me my cloud network topology.”

    Multi-cloud network topology visualizer. Prompt: ‘Show me my cloud network topology.’ Interactive diagram of EMEA / AMER / APAC regions, active providers, cross-connects. Bonus for live toggling MCN connections from the same UI.
Judged on · 4 criteria

  1. 01

    Does the demo run end-to-end?

    Live, no pre-recorded fakery.

    Required: 90-second demo video + GitHub repo

  2. 02

    Is emma essential to the demo?

    Couldn’t be staged on a single cloud.

  3. 03

    Is the idea sharp?

    Non-obvious move, not table stakes.

  4. 04

    Is it shareable?

    Worth posting without explanation.

02 / what you get

Everything you need to ship. Plus money if you do.

hardware

No contracts, no procurement, no sales calls. Spin up any GPU from a prompt.

prizes

Top 3 + People’s choice. Same rubric for builders and showcases.

// every submission — public gallery slot on hackathon.emma.ms, GitHub link in published projects, demo on our YouTube

03 / the process

From apply to winners.

Curation

Not gatekeeping. Show a shipped infra repo or a build thesis from our list to get in.

Credits

GPU and compute scaled to the scope of your approved build plan.

Sandbox

Limits set per approved plan. No infinite-spend traps.

  1. You can apply →
  2. Application reviewed
  3. Slack channel opens (#builders or #showcases)
  4. Submission deadline · firm
  5. Judging
  6. Winners announced
  7. Prizes wired (bank or crypto)

04 / judges

Invited industry judges.

05 / apply

We read every application.

Decisions land in your inbox within 2–3 business days. Approved teams get the credit pack and Slack invite the same day.

  1. 01 Apply ~5 min
  2. 02 Decision 2–3 days
  3. 03 Build 2–3 weeks / a weekend
  4. 04 Demo Live
  5. 05 Judging 21–25 Sep
choose your track
contact details

By submitting, you agree to our Terms and Privacy Policy.

06 / faq

Things you’re probably wondering.

If something else blocks you from applying — email .

What if my application is rejected?

We’ll tell you why. If it’s “we want to see shipped artifacts first” — pick something from the inspiration list, build it on your own, put it on GitHub, re-apply next time. We run these periodically.

What about my employer?

We don’t target companies. We target individuals. If your employer’s fine with you doing this on weekends, you’re fine.

What counts as “built on emma”?

Depends on the track. Builders: end-to-end through our API or SDK — provisioning, monitoring, maintenance. The full lifecycle. Showcases: provisioning and first run through our API or SDK. Monitoring/maintenance not required — a 90-second demo can’t fit a maintenance loop.

Is emma free for the hackathon?

For approved teams: yes, you get GPU and compute credits scaled to the scope of your approved build plan. For everyone else: no — emma is a paid platform. The credits are part of why we curate.

How much time does this realistically take?

Builders: 30–80 focused hours, usually spread over 2–3 weeks. Showcases: a long weekend (8–16 hours).

Ready?

First come, first served until the cohort fills.

Apply for hackathon