Michael Kowalski
Software Developer at Good Judgment Inc
Greater Brisbane Area Australia
Summary
🤩
RockstarMichael Kowalski is a software developer with 13 years of experience based in the Greater Brisbane Area, combining full-stack web development with machine learning engineering and test automation. At Roaring Deep he builds production web apps and automation—examples include federated GraphQL documentation tooling and a research-grade photo-based game built with React, Node.js and automated test suites. He contributes to notable open-source ML work, adding orthogonal initializers, QR decomposition and precision fixes to the swift-apis TensorFlow for Swift project while writing targeted unit tests for layers like BatchNorm and Conv2D. Michael also brings probabilistic reasoning skills from his role as a Superforecaster at Good Judgment, a background that sharpens his data-driven decision making and risk assessment. He blends practical engineering discipline with research-oriented rigor, comfortable shipping reliable code and verifying correctness through testing and numeric precision checks.
13 years of coding experience
Github Skills (13)
unit-testing10
swift10
machine-learning10
differentiable-programming10
deep-learning10
tensorflow10
normalization9
batch-normalization9
normalizing9
convolutional-neural-networks9
normalize9
test-automation8
gradient8
Programming languages (9)
Github contributions (5)
tensorflow/swift-apis
Jul 2019 - Jun 2020
Swift for TensorFlow Deep Learning Library
Role in this project:
ML Engineer & Test Automation Engineer
ML Engineer & Test Automation EngineerContributions:12 commits, 15 PRs, 48 comments in 11 months
Contributions summary:Michael primarily contributed to the `swift-apis` repository by implementing new features related to machine learning and deep learning, specifically focusing on the addition of orthogonal initializers and the QR decomposition functionality. They also worked on fixing the gradients for `rsqrt` function and ensuring that the precision matches in `gelu` function. Furthermore, the user wrote and updated unit tests for the implemented features and layers, including `BatchNorm`, `LayerNorm` and `Conv2D` layers, confirming the correct functionality, and also implemented tests for inferencing.
differentiable-programmingswift-for-tensorflowdeep-learningmachine-learningdeep-learning-library
mikowals/mojo
Dec 2023 - Aug 2024
The Mojo Programming Language
Contributions:43 pushes, 15 branches in 8 months
Find and Hire Top DevelopersWe’ve analyzed the programming source code of over 60 million software developers on GitHub and scored them by 50,000 skills. Sign-up on Prog,AI to search for software developers.