We gave away four NVIDIA Jetson Orin Nanos at Open Source Summit. Here's what people plan to build.

At Open Source Summit 2026 in Minneapolis, we had four NVIDIA Jetson Orin Nanos to give away. Instead of a raffle, we asked for something in return: a 60-second pitch on what you would actually build with one.

Before we get to the winners, here is what we loaded on every board that left our booth.

What is RLC Pro AI?

RLC Pro AI is CIQ's Enterprise Linux distribution built for AI, machine learning, and high-performance computing. It is a binary-compatible, downstream release built from RHEL source code, which means anything certified for Enterprise Linux runs on it without modification.

Out of the box, it ships with NVIDIA and AMD GPU drivers, NVIDIA CUDA, DOCA-OFED, and PyTorch preinstalled. The PyTorch flags, CUDA configurations, and kernel parameters are all tuned for throughput, and the goal was to get from installation to first inference in under four minutes. For data-heavy workloads, RDMA and InfiniBand support is included.

You get CIQ Portal access, up to four years of support including minor versions, per-node licensing, a secure supply chain, and deployment options across cloud, on-prem, and edge. For regulated environments, RLC Pro AI is certified for RHEL SPECs 5117, 5116, 5113, and 5095 across versions .2, .6, and .10 LTS.

Which brings us back to the Jetson Orin Nanos, and the four people who made the best case for walking out of Minneapolis with one.

The lecture that runs itself

One winner is a university employee working on an automated pipeline for faculty: record the lecture, get the teaching notes, get the transcript, done. Professors are already stretched. Pulling the post-production burden out of the equation is a real, practical time save, and a Jetson running local inference is a clean way to do it without sending everything to a cloud API.

Safety protocols nobody has reviewed since before the internet

Another winner works in an industry where some safety protocols have been sitting untouched for decades. Not because they are necessarily wrong, just because reviewing them at scale has never been practical. The plan is to use the board to build something that can finally surface, parse, and flag that documentation. That is the kind of unglamorous, high-stakes work that edge AI is genuinely good for.

Smaller models, specific jobs

One winner made a case for working against the grain. Rather than defaulting to a large general-purpose model, the plan is to build and deploy targeted SLMs tuned to specific software development tasks. Leaner, faster, and actually useful for the job at hand rather than everything at once.

A coach for navigating the social world

The pitch that stuck with us most came from a parent who wants to build a tool to help his neurodivergent son understand social contexts in real time. Not a replacement for human connection, just a patient, available resource that can help bridge the gap in moments that are hard to navigate. It was a good reminder that the most meaningful use cases for this hardware are not always the ones that show up in a product brief.

Try it yourself

You do not have to win a giveaway to get hands-on with RLC Pro AI. Head to portal.ciq.com and you can explore RLC Pro AI alongside the rest of the CIQ product line. If you are running AI workloads on edge hardware, on-prem clusters, or anywhere in between, it is worth a look.