about this event
Real AI. Real clusters. Built for SC26. Coming to Chicago for Supercomputing 2026? Start the week by building. The CIQ Supercomputing Local AI Hackathon brings together 20 teams for a deeply technical, hands-on AI build weekend immediately ahead of Supercomputing. Each team will get access to two Lenovo ThinkStation PGXs—powered by the NVIDIA GB10 Grace Blackwell Superchip—to use throughout the weekend, connected into a multi-node environment and orchestrated through Fuzzball. The challenge: build a real AI workload that actually takes advantage of the full cluster.
How to Participate: This is an intermediate, technical hackathon built for the AI and HPC community coming to Chicago for SC26, limited to a maximum of 20 teams. Once we reach capacity, additional applicants will be placed on a waitlist and notified if a spot becomes available. Application Guidelines: You must be able to attend the event in person. Technical Expertise: Experience working with multi-node systems and HPC environments. No prior Fuzzball experience is required; onboarding support will be provided. Team Size: Minimum 3 members, maximum 5 members. Priority: Full team applications will be prioritized over individual applications. We highly encourage you to form and apply as a team. Individuals: If you apply as an individual, team formation will happen the first day of the event. If accepted, you will be required to compete with your assigned team.
Challenge Tracks: Over one weekend, teams will build, benchmark, and demonstrate an AI or HPC workload running across their multi-node cluster. By Sunday, your team should be able to demonstrate both what you built and what the additional compute actually unlocked. 1. Multi-Node Training: Build a training or fine-tuning workload that takes advantage of compute across multiple nodes. 2. Inference at Scale: Explore multi-node inference for greater throughput, concurrency, model capacity, or performance. 3. Scientific AI: Apply distributed AI and accelerated computing to a scientific or research workload. 4. Cluster & Scheduling: Build around orchestration, scheduling, resource allocation, utilization, or multi-node infrastructure.
What You Can Win: We're sending the top 3 teams straight from the hackathon into Supercomputing 2026. Prizes include SC26 conference passes, a Lenovo ThinkStation PGX ($6,000 ABV), an official prize moment and team photo at the CIQ x NVIDIA booth during SC26, and an NVIDIA Developer Livestream feature.
Preparation Resources: Start Building on DGX Spark — find instructions and examples to run AI workloads on the NVIDIA GB10 Grace Blackwell Superchip. NVIDIA Build Model Endpoints — access low-latency, high-accuracy endpoints designed for agentic systems. Fuzzball — understand how to set up an AI model in v4.3 of Fuzzball. Privacy policies referenced from CIQ, NVIDIA, VAST Data, Lenovo, and Drive Capital. Thank you to our sponsors.
tracks
Multi-Node Training
Build a training or fine-tuning workload that takes advantage of compute across multiple nodes.
Inference at Scale
Explore multi-node inference for greater throughput, concurrency, model capacity, or performance.
Scientific AI
Apply distributed AI and accelerated computing to a scientific or research workload.
Cluster & Scheduling
Build around orchestration, scheduling, resource allocation, utilization, or multi-node infrastructure.
when
CIQ Supercomputing Local AI Hackathon
Day one: team formation for individual applicants, onboarding, and kickoff of the multi-node AI build weekend.
◍ TBAtickets
| tickets | price | Fee | Status | description |
|---|---|---|---|---|
| Free Registration | Free | TBA | Available | Free to attend; applications are limited to a maximum of 20 teams, with additional applicants placed on a waitlist. |
prizes
| # | PRIZE | price |
|---|---|---|
| Top 3 Teams | SC26 conference passes; Lenovo ThinkStation PGX ($6,000 ABV); official prize moment and team photo at the CIQ x NVIDIA booth during SC26; NVIDIA Developer Livestream feature | $6,000 ABV |