Articles

• A Recipe For AI Strategy (ODSC Featured Article)
• Fooling Fintech AIs (Grow Blog)
  • Part I - Introduction
  • Part II - How to Fool the Machines
  • Part III - Protecting your Business from Adversarial Users
• Agile For Research
• Collaborating on Data Science Models
• Statement of Data Science Philosophy
• Guide to running your own web services

Presentations

• Improving our Teams' Learning Cultures + slides (2024-02-23)
• Machine Learning and AI for Interstellar Colonization (2021-05-26)
• Kaggle Meetup: A Recipe for Training Neural Networks (2019-06-14)
• Kaggle Meetup: Quora Insincere Questions Classification (2019-03-08)
• Python for high performance and scientific computing (2013-04-07)

Open Source

• Links Clustering - Python implementation of the Links online clustering algorithm. Unlike most clustering algorithms, this one predicts cluster membership as it goes without needing to see the entire data set. Featured on paperswithcode.com.
• DLCheck - A tool for automatically discovering training problems in PyTorch Lightning Deep Learning models. Implements checks from TFCheck and Deep Diagnosis.
• hwc - The Hello World Compiler - A compiler that takes input source code written in any programming language and compiles it into a working program with the desired features. It is useful for accelerating development of new programming languages, or for writing software using languages you don't understand (yes, it's a joke).
• Bug fixes and features
  • Librosa - Deemphasis - Implemented the deemphasis effect and added unit tests in a major audio library (6.5K stars).
  • TSFresh - Index Preservation - Found and fixed a bug involving incorrect handling of indices and added unit tests in a major time-series library (8k stars).

Contact

• LinkedIn
• Twitter
• Github
• Mastodon