Working ahead of the frontier
What it takes to push models past their out-of-the-box capabilities and stay six months or more ahead of them. Drawn from work across Gemini, Gemma, WeatherNext, MetNet and more.
Speaking
Speaker kit
I speak about applied AI, API design and payments infrastructure. If you're putting together an event, everything you should need is on this page. If something's missing, email me and I'll get back to you.
Bio
JJ Geewax is Director, Applied AI at Google DeepMind, where he pushes the technical boundaries of what DeepMind's models can do. He is the author of API Design Patterns and Google Cloud Platform in Action, and the founder of AIP.dev, the API design guidance used across Google.
JJ Geewax is Director, Applied AI at Google DeepMind in Singapore, where his team pushes the technical boundaries of DeepMind's models, from Gemini and Gemma to WeatherNext and MetNet, staying six months or more ahead of their out-of-the-box capabilities. Before DeepMind he was at Meta, working on WhatsApp Business and Ad Signal and serving as a technical advisor to the COO. He spent more than a decade at Google before that, leading API design across Alphabet, founding AIP.dev, and working with central banks and regulators on real-time payment systems. JJ is the author of API Design Patterns and Google Cloud Platform in Action, and holds a degree in computer science from the University of Pennsylvania.
JJ Geewax is Director, Applied AI at Google DeepMind, based in Singapore. His team pushes the technical boundaries of DeepMind's models, working across Gemini, Gemma, WeatherNext, MetNet and more. Everything they build aims to stay six months or more ahead of the models' out-of-the-box capabilities.
Before joining DeepMind, JJ was at Meta. He worked on WhatsApp Business, including AI adoption, WhatsApp Flows and developer tools, and on Ad Signal, Meta's work on Pixel and behavioral signals for ad targeting. He then served as a technical advisor to the COO, exploring new AI-driven business lines in Asia.
JJ spent more than a decade at Google. On the Payments team he led Google's contributions to Mojaloop, an open-source real-time payment system, represented Google on the Mojaloop Foundation's Technical Governing Board, and advised central banks and regulators on real-time payments and third-party payment initiation. On Google Cloud he led the API design team across Alphabet and founded AIP.dev, the API Improvement Proposals that describe how Google designs its APIs. He joined Google when it acquired Invite Media, where he was VP of Engineering, and went on to lead east coast engineering for DoubleClick Bid Manager. He has also served on the BIS Innovation Hub's Project Rosalind API advisory group and the board of advisors of Singapore's Open Government Products.
JJ is the author of API Design Patterns (Manning, 2021) and Google Cloud Platform in Action (Manning, 2018), and has published research on API governance (ICSE 2024) and AI-assisted API design (CHI 2026). He has a Bachelor's degree in computer science from the University of Pennsylvania and lives in Singapore with his wife and son.
Topics
AI-assisted API design
When a model trained on Google's API guidelines writes specs that score better than most human designers, what's left for the designer to do? Based on our CHI 2026 study.
API design and governance at scale
How Google keeps its APIs consistent across the whole company, from API Improvement Proposals to automated linting and readability reviews. Drawn from API Design Patterns and our ICSE 2024 paper.
Real-time payments and financial inclusion
Open-source payment infrastructure, third-party payment initiation, and lessons from working with central banks and regulators on national payment systems.