JAX for HPC/AI Hybridization Hackathon
September 7–9, 2026 • IDRIS, Paris-Saclay
The JAX for HPC/AI Hybridization Hackathon is a three-day hands-on event dedicated to evaluating JAX for large-scale scientific computing and HPC/AI convergence. Participants will explore JAX’s strengths and limitations through practical benchmarks, scientific computing applications, and scalability studies on the Jean Zay supercomputer.
Two participation tracks are offered:
Track 1 – JAX for Scientific Computing (Beginner Track)
Designed for researchers and engineers with experience in scientific computing, HPC, or numerical methods who are new to JAX.
Participants will:
- Learn the fundamentals of JAX through daily crash courses.
- Apply newly acquired skills to scientific computing case studies.
- Build reproducible benchmarks comparing JAX with established scientific computing frameworks.
Track 2 – Pushing the Limits of JAX (Advanced Track)
Designed for experienced JAX users interested in large-scale performance and scalability.
Participants will:
- Work on real-world JAX applications throughout the event.
- Investigate multi-GPU and multi-node scaling on Jean Zay.
- Identify bottlenecks, missing features, and optimization opportunities.
- Produce advanced benchmarking and scalability studies.
Agenda
|
Time |
Monday, Sept. 7 |
Tuesday, Sept. 8 |
Wednesday, Sept. 9 |
|
|
|||
|
09:30 – 10:30 |
|
Track 1: JAX crash course |
Track 1 / Track 2: Practical work |
|
10:30 – 11:00 |
Welcome coffee |
Coffee break |
Coffee break |
|
11:00 – 12:30 |
Track 1: JAX crash course - Rémi Imbach |
Track 1: JAX crash course - E.Franck |
Track 1 / Track 2: Practical work |
|
12:30 – 14:00 |
Lunch |
Lunch |
Lunch |
|
14:00 – 15:30 |
Track 1: Hands-on scientific computing project |
Track 1: Hands-on scientific computing project |
Track 1/ Practical work |
|
15:30 – 16:00 |
Coffee break |
Coffee break |
Coffee break |
|
16:00 – 18:00 |
Track 1: Hands-on scientific computing project |
Track 1: Hands-on scientific computing project |
Common wrap-up session, presentations & conclusions |
Dinner on tuesday at 19:30 - more information to come
Practical Information
Dates: September 7–9, 2026
Venue: IDRIS, Paris-Saclay
Computing resources: Jean Zay supercomputer
Format: In-person
Tracks: Beginner and Advanced
Three days to learn, benchmark, and scale JAX for next-generation HPC and AI workloads !
Organized by

