Skip to main content
Sticky

The Prompt Project Challenge - How It Works

  • July 9, 2026
  • 9 replies
  • 1662 views
Axelera AI's The Prompt Project Challenge
Spanner
Axelera Team
Forum|alt.badge.img+3

✈️ Every Maverick Needs a Wingman

Voyager Wingman is Axelera's new AI pipeline generator. You describe what you want to build, and it generates a working Metis pipeline, runs it on the hardware, and iterates until it works. The YAML and the wiring are its job. The awesome project ideas are yours.

The Prompt is our community challenge to see how far you can take that. Same hardware, same SDK, a very different starting point.

 

🎯 The Mission

Pitch us an idea for an AI project you want to build using Voyager Wingman prompts on Axelera’s powerhouse AI system, the Dell Pro Slim Plus XE5 with a 16GB Axelera PCIe card. We’ll pick the ten most inspiring, creative project ideas, and send each selected participant a Dell XE5, and set you loose for a one-month build.

Take it out of the box, get Voyager Wingman on board (a quick install straight from the web chat), and you’re building on real Axelera silicon the same day. 🚀

 

Not Selected? Build Anyway!

Voyager Wingman is free and open to everyone, not just the ten project proposals we pick. So if you already have Axelera hardware, you’re welcome to join in with the challenge!

Build a working project with Voyager Wingman, meet the same submission criteria by the same deadline, and you’re in the community category. No pitch, no selection, no kit from us; just build and post. The best community submission wins an all-new Axelera AI Mini PC with an Axelera Embedded 113m on board! 🖥️

 

What We’re Looking For

The same for everyone, selected or not, your final project submission must include:

  • The prompt itself, shared in full, so the community can see where you started.

  • A working demo running on your Axelera system. Not "it would work if...". We want to see it live and in action.

  • Your journey: The Voyager Wingman prompts you used, any iterations made with Voyager Wingman, all updates, photos, videos and code repos.

It doesn’t all have to run from the very first collaboration with Voyager Wingman, of course. It’s a journey, like any AI project. But we want you to show us how fast and how impressive Voyager Wingman is by building a project in record time. Let Wingman do the heavy lifting, while you bring the creative building.

 

The Need for Speed

Since Voyager Wingman throws the AI project building engines into overdrive, if you’ve finished the project you proposed ahead of schedule, why not create more? Post as many additional projects as you want to increase your chances of winning! If you’ve finished and submitted your original proposal, that is…

 

🏆 Prizes

For the ten selected builders:

  • Dell Pro Slim Plus XE5 with a 16GB Axelera PCIe card to build your Wingman project on, and to keep.

  • Free access to Voyager Wingman.

  • 💶 €1,000 for the winning project, and €500 for runner-up.

Community category, open to everyone:

  • Enter using your own Axelera hardware and Voyager Wingman, with same criteria and deadline.

  • The best community submission wins an Axelera AI Mini PC.

And every valid submission, selected or community, gets a permanent spot in the community's Prompt gallery, with full credit. 🌟

 

🗓️ How it Runs

  • Pitches open: 15 July

  • Pitches close: 09 August (extended from 29 July)

  • Selected builders announced and kits shipped: Early August

  • Build phase: one month (tight by design! You’ve got a wingman on this build, after all)

  • Final project submissions (selected and community): 13 September

  • Winners announced: Week of 21 September

 

✍️ How to Enter

To be one of the ten: Post your project proposal right here in The Prompt challenge group in its own brand new thread, then fill in the short form on our website with your details (these details are only used to help with provision of the Dell XE5 kit if you’re selected, nothing else).

Community category: Already have your own Axelera hardware? Build your project and post it in the group before the final deadline, with your prompt, working demo and repo(s).

Go be a maverick!

9 replies

Ailean
  • July 15, 2026

Hi, 
When I try and fill in the  short form on our website, after pressing [Submit] I get send to a standard google.com Google Search page. Is this correct?

Google https://www.google.com/?submissionGuid=2fb3289d-4d17-4a8d-b888-9128a0c0cc65

The information before the “?” is missing and therefore will send me to the google.com search website. 

Can you fix this somehow as I have already sent in my The Prompt proposal?

Thanks in advance,
Ailean


Spanner
Axelera Team
Forum|alt.badge.img+3
  • Author
  • Axelera Team
  • July 16, 2026

Hi, 
When I try and fill in the  short form on our website, after pressing [Submit] I get send to a standard google.com Google Search page. Is this correct?

Google https://www.google.com/?submissionGuid=2fb3289d-4d17-4a8d-b888-9128a0c0cc65

The information before the “?” is missing and therefore will send me to the google.com search website. 

Can you fix this somehow as I have already sent in my The Prompt proposal?

Thanks in advance,
Ailean

Hmm, let me look into that today and get back to you. Don’t worry though, we’ll make sure it’s working well before we hit any deadlines. The post on here and the form don’t have to be done at exactly the same time 👍

Actually, it doesn’t hurt to clarify - the form is just used to make sure that we can deliver both equipment and gear to people who are joining in with The Prompt, but without asking people to make that info public! It’s not used for anything else.


Ailean
  • July 16, 2026

Hi,

Thank you for checking, and also for clarifying what the form is used for. That makes sense, especially as the delivery details should remain private.

I’m glad to know that my proposal post has already been received and that the form can be completed separately once the issue is fixed. I’ll wait for your update and try again when the form is working properly.

Thanks again for your help and reassurance.

Best wishes,
Ailean


Spanner
Axelera Team
Forum|alt.badge.img+3
  • Author
  • Axelera Team
  • July 16, 2026

Just a quick note, guys - I haven’t had a chance to dig deep into the mechanics of the form on the website yet, but I did just notice that the forms are coming through to us anyway. So we’re all good in terms of your forms being submitted, even if it does something a bit weird afterwards 😄


Ailean
  • July 16, 2026

Ok that's great then. My apologies for the many forms you'll receive from me as I tried it plenty of times. Can you tell me if the form just consists of the one page? As that's how far I came, there isn't a second page where you have to fill in your address or something? Thanks in advance! I'll lay of filling the form in again then lol.


Spanner
Axelera Team
Forum|alt.badge.img+3
  • Author
  • Axelera Team
  • July 16, 2026

No worries at all, it’s good to have given it a full field test! 😄

It’s just one page, yep, the six fields you can see here. I’ll get in touch for full delivery details for those projects that get selected to go forward. But for this moment, the info below is plenty.

 


Forum|alt.badge.img
  • Cadet
  • August 3, 2026

the form submission problem persists - I tried submitting today and it redirected to google home page.


I am still having problems with the form submission, and I am getting redirected to the Google Home Page.


Object Memory — An Edge AI That Learns How Your Space Behaves

I want to build a small edge-AI system that learns how objects and people normally interact with a space.

The idea is simple: a camera watches a desk or room and detects objects, people and their spatial relationships. Instead of only classifying what is visible, the system builds short temporal sequences such as:

person → cup → lifted → moved → table

or

person → keys → picked up → door.

Over time, it learns which interactions and sequences are normal. When something happens that does not fit what it has previously observed, it flags it as an unusual event.

The interesting part is that I don't want to train a separate classifier for every possible human action. I want to experiment with representing the scene as a small interaction graph and learning the temporal patterns between objects.

Everything would run locally on the Axelera Metis card.

I can build and test the entire prototype at home with a normal webcam, a desk and everyday objects, which makes the project realistic to complete within one month.

The question I want to explore is:

Can a small edge-AI system learn what normally happens in a space, rather than simply recognizing what is inside it?