Andrew Trask
Privacy-Preserving AI Researcher
Term Member at Council on Foreign Relations.
Senior Research Scientist at Google DeepMind, studying privacy and AI: federated learning, differential privacy, and secure computation.
PhD student (ABD) at University of Oxford, focusing on privacy-preserving ML. Funded by the Oxford-DeepMind Graduate Scholarship. Affiliated with FHI and GovAI.
Wrote Grokking Deep Learning (Manning). Also teach in Udacity's Deep Learning Nanodegree.
Researcher at Digital Reasoning, training the world's largest neural network (160B params). ICML.
BS in Applied Discrete Mathematics and BBA in Finance at Belmont University.
I'm the founder of OpenMined, working on technologies to increase the capability of and deconcentrate power over AI by combining deep learning, cryptography, and distributed systems. I'm particularly interested in attribution-based control for AI systems and broad listening. PhD student at Oxford. Research Scientist at Google DeepMind. Term Member at the CFR.
publications
- 2025 Beyond Release: Access Considerations for Generative AI Systems,arXiv
- 2025 Private Map-Secure Reduce: Infrastructure for Efficient AI Data Markets,arXiv
- 2025 Robust AI Personalization Will Require a Human Context Protocol,SSRN
- 2025 Now is the time for Privacy Enhancing Technologies,Statistical Journal of the IAOS
- 2024 Open Problems in Technical AI Governance,arXiv
- 2024 The Ethics of Advanced AI Assistants,arXiv
- 2024 Personhood Credentials: AI and the Value of Privacy-Preserving Tools to Distinguish Who is Real Online
- 2023 UN Handbook on Privacy-Preserving Computation Techniques,United Nations
- 2023 Exploring the Relevance of Data Privacy-Enhancing Technologies for AI Governance Use Cases,arXiv
- 2022 Artificial Intelligence in Medicine and Privacy Preservation,Book chapter
- 2021 End-to-end Privacy Preserving Deep Learning on Multi-institutional Medical Imaging,Nature MI
- 2021 PySyft: A Library for Easy Federated Learning,Springer
- 2021 DP-SGD vs PATE: Which Has Less Disparate Impact on Model Accuracy?,arXiv
- 2021 Syft 0.5: A Platform for Universally Deployable Structured Transparency,arXiv
- 2021 Sensitivity Analysis in Differentially Private ML Using Hybrid Automatic Differentiation,arXiv
- 2021 An Automatic Differentiation System for the Age of Differential Privacy,arXiv
- 2021 Towards General-purpose Infrastructure for Protecting Scientific Data Under Study,PPML Workshop
- 2020 The Future of Digital Health with Federated Learning,NPJ Digital Medicine
- 2020 Toward Trustworthy AI Development: Mechanisms for Supporting Verifiable Claims,arXiv
- 2020 Neither Private Nor Fair: Impact of Data Imbalance on Utility and Fairness in DP,PPML
- 2020 Beyond Privacy Trade-offs with Structured Transparency,arXiv
- 2020 Privacy-Preserving Medical Image Analysis,arXiv
- 2020 Benchmarking Differentially Private Residual Networks for Medical Imagery,arXiv
- 2019 Systems and Methods for Neural Language Modeling,US Patent
- 2018 Neural Arithmetic Logic Units,NeurIPS
- 2018 A Generic Framework for Privacy Preserving Deep Learning,arXiv
- 2018 Sample Efficient Adaptive Text-to-Speech,arXiv
- 2018 Scaling Shared Model Governance via Model Splitting,arXiv
- 2015 sense2vec: Fast and Accurate Word Sense Disambiguation in Neural Embeddings,arXiv
- 2015 Modeling Order in Neural Word Embeddings at Scale,ICML (Oral)
teaching
Teaches sentiment analysis with neural networks for text processing.
18,000+ studentsPrivacy infrastructure and structured transparency. Free with certificate.
~7 hoursHomomorphic encryption, MPC, and differential privacy.
FreeCross-organizational federated learning systems.
Freefeatured talks
book
Grokking Deep Learning
Learn to build neural networks from scratch using only Python and NumPy. No frameworks, no magic—just the fundamentals.
blog posts
From OpenMined and iamtrask.github.io
- 2025-10-19 GPU demand is (~1Mx) distorted by efficiency problems which are being solved
- 2025-10-06 Unlocking a Million Times More Data for AI Through Attribution-Based Control
- 2025-10-06 The Bitter Lesson's Bitter Lesson
- 2025-09-01 What is Broad Listening?
- 2023-11-21 Written Statement for the U.S. Senate AI Insight Forum
- 2022-01-20 Announcing our partnership with Twitter to advance algorithmic transparency
- 2020-03-16 Apply to Join the Syft Core Team
- 2020-03-05 OpenMined + apheris AI Partnership for PyTorch Mobile Federated Learning
- 2020-03-05 Introducing OpenMined Research
- 2019-12-06 OpenMined and PyTorch Join Forces to Advance Privacy-Preserving Machine Learning
- 2017-06-19 Is the Universe Random?
- 2017-06-05 Safe Crime Detection
- 2017-03-21 Deep Learning without Backpropagation
- 2017-03-17 Building Safe A.I.
- 2017-01-15 Tutorial: Deep Learning in PyTorch
- 2016-08-17 Grokking Deep Learning
- 2016-02-25 How to Code and Understand DeepMind's Neural Stack Machine
- 2015-11-15 Anyone Can Learn To Code an LSTM-RNN in Python (Part 1: RNN)
- 2015-07-28 Hinton's Dropout in 3 Lines of Python
- 2015-07-27 A Neural Network in 13 lines of Python (Part 2 - Gradient Descent)
- 2015-07-12 A Neural Network in 11 lines of Python (Part 1)
- 2014-11-23 Word2Vec Analysis of Harry Potter
- 2014-11-23 Tutorial: when Numpy isn't fast enough...
- 2014-11-22 Tutorial: Spark-GPU Cluster Dev in a Notebook