pearu / pytorch

Tensors and Dynamic neural networks in Python with strong GPU acceleration

• I'm currently a Staff Software Engineer at Quansight

• My current interests in Scientific Computing include (see here for technical blogs):

  • Validation of mathematical functions with complex inputs in various libraries such as NumPy, PyTorch, XLA/JAX, Tensorflow, MPMath, etc
  • Theory of Sparse Arrays, in particular, PyTorch sparse tensors support
  • Optimization of Triton kernel parameters
  • Interoperability of Array-like objects from various software
  • Algorithms and Implementations
  • Support Fortran 90 TYPE construct in F2Py

• I have initiated and co-authored a number of open source projects with the following highlights:

  • RBC - Remote Backend Compiler
  • F2Py - Fortran to Python interface generator DOI, ships with NumPy DOI
  • SciPy - open-source software for mathematics, science, and engineering DOI

  and contributed to a number of open source projects such as XLA, JAX, Heavy.AI, PyTorch, Numba, Apache Arrow, conda-forge, XND, SymPy, etc.

• My background is in Scientific Research: Mathematical Physics, Nonlinear Waves, Mechanics, Systems Biology, Microscopy, Solid Mechanics, Optics, Scientific Computations, etc. CV@ERIS, ORCID, Scholar

• 📫 How to reach me: Linkedin, GMail

• Fun facts:

  • I have completed AoC 2019, AoC 2020
  • Estonian Champion 2006 in Dog Obedience (German Shepherd), currently training for Agility (Boston Terrier)
  • Ship engineer by profession, raced with power boats (Circuit S-250, S-350, OSY-400, O-125, F-500), teached sailing to kids

• 😄 Pronouns: standard Modern English