John Aslanides
Staff Research Engineer at DeepMind
London, England, United Kingdom
Summary
🤩
Rockstar🎓
Top SchoolJohn Aslanides is a Staff Research Engineer at DeepMind in London with 10 years of experience building production-grade machine learning and AI systems. He is an active open-source contributor to heavyweight projects such as Acme, Haiku, JAX and TensorFlow Probability, focusing on reliability (checkpointing and R2D2 learner improvements), clearer Transformer/MNIST examples, and code quality via type annotations. His career bridges research and engineering—from gravitational-wave data analysis to deploying scalable RL components—anchored by a University Medal in Computer Science from ANU. He’s known for pragmatic backend and DevOps changes that reduce training brittleness while improving the developer experience. Outside core ML work he has taught theoretical physics and even classical piano, which surfaces in his meticulous documentation and example-driven contributions.
10 years of coding experience
8 years of employment as a software developer
High School, NSW HSC, 99.25 UAI, High School, NSW HSC, 99.25 UAI at Canberra Grammar SchoolAustralian National University
English, French, Spanish
Stackoverflow
137reputation
16kreached
0answers
4questions
Github Skills (37)
transformers10
probabilistic-programming10
checkpoint10
python10
machine-learning10
reinforcement-learning10
numpy10
haiku10
checkpointing10
agents10
type-annotations10
tensorflow210
agent10
deep-learning10
tensorflow10
Programming languages (7)
Github contributions (5)
google-deepmind/bsuite
Aug 2019 - Oct 2021
bsuite is a collection of carefully-designed experiments that investigate core capabilities of a reinforcement learning (RL) agent
Role in this project:
Data Scientist
Data ScientistContributions:7 releases, 1 review, 96 commits in 2 years 2 months
Contributions summary:John removed an unused section from an analysis notebook related to how to cite the project, indicating a focus on refining the presentation of results. This suggests a role in data analysis, ensuring the clarity and conciseness of project documentation. Their work is likely aimed at improving the user experience by cleaning up the notebook and removing unnecessary content.
stable-baselinesagentreinforcement-learningdeep-reinforcement-learningcapabilities
google-deepmind/acme
May 2020 - Aug 2021
A library of reinforcement learning components and agents
Role in this project:Back-end Developer & DevOps Engineer
Back-end Developer & DevOps EngineerContributions:4 releases, 83 commits, 1 PR in 1 year 3 months
Contributions summary:John primarily contributed to the core components of the reinforcement learning library. Their work involved significant modifications to the agent's checkpointing and model saving mechanisms, improving the stability and efficiency of the training process. They refactored the code to make checkpointing an optional feature in agents like R2D3, and made enhancements to the R2D2 learner. They also added a global acme_id flag for finer-grained control over path creation.
reinforcement-learningreinforcementagentsdeep-reinforcement-learning
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.