Join our team at gradient for the "Build your own AI lab" competition! Starting today, participants are challenged to build an application or workflow using the local inference capability of parallax to run your own AI.
We’re running this campaign in two tracks:
Track 1 – Showing Cluster Setups
Show us your best Parallax-powered local AI cluster in the wild — from cozy home rigs to overkill racks. Post a photo of your desk or cluster setup running Parallax and tag @Gradient_HQ on X. An example would look like this:
Our Parallax cluster setup
Track 2 – Building Applications
Show us what you’re building with Parallax — personal assistants, agents, tools, workflows, anything that pushes what local AI can do. Share a short demo, repo, or thread walking through your app and how Parallax powers it.
To learn more and get questions answered, join the parallax Discord.
Submissions are open now and will be closed November.30.2025 (11:59pm US - Eastern Standard Time).
Share your project with the community on X, Reddit, and in our Discord! Tag gradientHQ on X to be sure we see your work. Top projects will have the opportunity to win awesome hardware that can be used to run large models locally!
- 1st place - DGX Spark
- 2nd - 8th place - Mac-mini
After the judges have made their picks, the DGX Spark and one Mac mini will be provided by Ahmad in collaboration with the team at gradient. Give Ahmad a shoutout on X! The remaining Mac minis will be provided by the gradient team.
Judging Criteria
The world needs useful, local AI applications that we can rely on to be private and low cost! Our judges (stay tuned for more) will be looking for awesome setups, impactful applications that solve problems for people and businesses. Applications will be reviewed as they are submitted and winners will be announced on December.02.2025. The primary factors that will determine the winners are
- Track 1: Share your most creative setup using Parallax. Post about the project across social media!
- Track 2: Share your most useful application built locally. Solve problems for individuals that require privacy and low cost. Help researchers accelerate their work with local tooling that is easy to use. Build tools for businesses that create efficiencies.
How to Participate
Timeline: November.17.2025 - November.30.2025
For Track 1, simply share your awesome cluster setup and @Gradient_HQ on X!
For Track 2, submit your application/agent to the form here.
Must include:
- Provide at least one post (X, Reddit, RedNote, etc.) with screenshots of the parallax use case and a description of the application. The more you post, the higher your likelihood of winning and providing impact to the community.
- Tell us how you used parallax and why in your social posts. The more detail the better! Help us teach the community about building local AI applications.
Info that improves your chances of winning:
- Github repo
- Video of the application
- Social posts supporting the campaign
More on parallax
Hosting AI models locally is important for user control over the AI tooling we use everyday. With Parallax, users can stitch together consumer hardware like Apple computers and desktops with Nvidia GPUs to host large models locally. No need for expensive enterprise grade GPUs and your data is kept local and private.
Applications built using this local inference endpoint are user owned. For businesses, hosting an on-prem AI solution not only cuts costs, but improves efficiencies in an AI powered system.
- Reduce your cost of operations - no Inference API costs, no paid subscriptions
- Data inputs and outputs are private
- Ownership - You control the model version
Parallax is the world's first fully distributed serving framework that turns a pool of heterogenous GPUs, despite their locations, into an efficient inference platform. Get complete access to the state-of-the-art open source models hosted on distributed devices collectively, for free!
Deploying a large language model (LLM) inference service remains costly because centralized serving depends on specialized GPU clusters and high-bandwidth interconnects in data centers. An appealing alternative is to leverage collaborative decentralized GPU pools. However, heterogeneity in GPU and limited interconnected network bandwidth, along with potentially dynamic availability, make efficient scheduling the central challenge in this scenario.
How Parallax works under the hood
Parallax addresses this problem with a decentralized LLM serving system that turns a pool of heterogeneous GPUs into an efficient inference platform via a two-phase scheduler. Parallax decomposes planning into (i) model allocation, which places layers of each replica across diverse GPUs to jointly optimize latency and throughput under memory and link-bandwidth constraints, and (ii) request-time GPU pipeline selection, which stitches layers from different replicas into end-to-end execution chains that balance load and adapt to current conditions.
Check out Parallax on Github and read more in the white paper!
Learn more about Gradient and our mission of building the open intelligence stack.