Description: Personal website of Jason Ansel.
Jason Ansel
I am currently working on PyTorch at Meta where I started the TorchDynamo and TorchInductor projects, both of which are foundational parts of PyTorch 2.0 . TorchDynamo is a Python-level JIT compiler designed to make unmodified PyTorch programs faster. TorchDynamo hooks into the frame evaluation API in CPython (PEP 523) to dynamically modify Python bytecode right before it is executed. TorchInductor is a new compier backend for TorchDynamo that maps arbitrary PyTorch programs to Triton or C++/OpenMP.
Before Meta, I was at GoDaddy helping build a deep learning platform for predicting small business behavior and personalizing experiences across the company. I also created GoDaddy Domain Appraisals , which uses neural networks to predict the resale value of a domain name better than a human expert. I joined GoDaddy in 2013 as part of the acquisition of the startup Locu , which I joined in 2011 while I was simultaneously getting my Ph.D. at MIT CSAIL .