brain of mat kelcey...

keras3 to jax to tflite/liteRT via orbax

yolz; you only look o̶n̶c̶e̶ zero times

differentiable kalman filters in jax

a (larger) wavenet neural net running on an FPGA at (almost) 200,000 inferences / sec

a wavenet neural net running on a microcontroller at (almost) 50,000 inferences / sec

high performance ML with JAX

evolved channel selection

crazy large batch sizes

solving y=mx+b... with jax on a tpu pod slice

develomentor.com podcast interview