# JAX Projects at AI Tinkerers

> Canonical HTML: https://aitinkerers.org/technologies/jax
> Markdown URL: https://aitinkerers.org/technologies/jax.md
> Technology record last updated: 2026-03-31T11:48:33Z
> Generated: 2026-09-20T22:40:59Z

JAX combines Autograd and XLA to deliver high-performance numerical computing and machine learning research at scale.

JAX transforms NumPy code into hardware-accelerated kernels using a functional API. It leverages XLA (Accelerated Linear Algebra) to target GPUs and TPUs, achieving massive throughput for deep learning and scientific simulations. Key primitives like jit (just-in-time compilation), vmap (automatic vectorization), and grad (arbitrary-order differentiation) allow developers to write pure Python while executing at native speeds. By treating programs as composable transformations, JAX eliminates the overhead typical of standard Python execution and provides a unified framework for modern AI research.

- Official technology site: https://github.com/google/jax
- Public AI Tinkerers demos and talks: 2
- Result page: 1 of 1

## Recent Public Talks and Demos

### [How do you train AlphaZero-style RL models at 100M steps/sec on one GPU using Jaxpot](https://poland.aitinkerers.org/talks/rsvp_M0hzwiMQakQ)

We built bards-ai/Jaxpot: an open-source JAX library for training RL agents via self-play, built around extreme environment parallelism. Everything runs on the GPU - env steps, agent forward passes, and training - so you can hit 100M+ environment steps/second on a single GPU - a peace that makes RL training viable on consumer-level hardware. I'll show live how to code an environment in a GPU-parallelism-friendly way, how to set up agents, and how to train an expert-level agent with zero supervised data - purely by models playing themselves in environment, AlphaZero-style.

- Event context: AI Tinkerers Poland #8 - Meetup in Warsaw (18th June) — 2026-06-18 — Poland
- Public talk page: https://poland.aitinkerers.org/talks/rsvp_M0hzwiMQakQ

### [CyberRunner: How We Open-Sourced the AI That Beat Humans](https://zurich.aitinkerers.org/talks/rsvp_95B2B3fk6Sc)

CyberRunner is the first AI-driven robotic system to learn and master the physical "Labyrinth" marble game through reinforcement learning, achieving times that surpass the world record set by humans. This talk explores the transition from a laboratory experiment to a fully open-sourced hardware and software stack designed to democratize high-speed, high-precision robotic control.

- Event context: AI Tinkerers Zurich April 9th — 2026-04-09 — Zürich
- Public talk page: https://zurich.aitinkerers.org/talks/rsvp_95B2B3fk6Sc

## Related Technologies

- [bards-ai/jaxpot](https://aitinkerers.org/technologies/bards-ai-jaxpot) ([Markdown](https://aitinkerers.org/technologies/bards-ai-jaxpot.md)) — 1 public demo
- [DreamerV3](https://aitinkerers.org/technologies/dreamerv3) ([Markdown](https://aitinkerers.org/technologies/dreamerv3.md)) — 1 public demo
- [Flax](https://aitinkerers.org/technologies/flax) ([Markdown](https://aitinkerers.org/technologies/flax.md)) — 1 public demo
- [Hydra](https://aitinkerers.org/technologies/hydra) ([Markdown](https://aitinkerers.org/technologies/hydra.md)) — 1 public demo
- [jaxpot](https://aitinkerers.org/technologies/jaxpot) ([Markdown](https://aitinkerers.org/technologies/jaxpot.md)) — 1 public demo
- [OpenCV](https://aitinkerers.org/technologies/opencv) ([Markdown](https://aitinkerers.org/technologies/opencv.md)) — 26 public demos
- [pgx](https://aitinkerers.org/technologies/pgx) ([Markdown](https://aitinkerers.org/technologies/pgx.md)) — 1 public demo
- [Python](https://aitinkerers.org/technologies/python) ([Markdown](https://aitinkerers.org/technologies/python.md)) — 662 public demos
- [TensorFlow](https://aitinkerers.org/technologies/tensorflow) ([Markdown](https://aitinkerers.org/technologies/tensorflow.md)) — 90 public demos