David Ha is a Canadian machine learning researcher and entrepreneur. He is the co-founder and chief executive officer of Sakana AI, a Tokyo-based artificial intelligence startup. Ha is known for his work on neural networks, including research on attention mechanisms and generative models, and for his leadership in establishing Google Brain's Tokyo lab.
Born in British Hong Kong, Ha pursued higher education in Canada and Japan. He earned an undergraduate degree in engineering science and a master's degree in mathematical finance from the University of Toronto, followed by a PhD from the University of Tokyo. Before entering the field of AI research, Ha worked as a derivatives trader, eventually becoming a managing director at Goldman Sachs in Japan.
Early Career in Finance
Ha's career began in quantitative finance. After completing his master's degree, he joined Goldman Sachs in Tokyo, where he worked as a derivatives trader. Over time, he rose to the position of managing director, a senior role that involved overseeing trading strategies and risk management. His experience in finance would later inform his approach to machine learning, particularly in areas like reinforcement learning and decision-making under uncertainty.
Despite his success in finance, Ha was drawn to the emerging field of deep learning. He began exploring neural networks as a hobby, studying online resources and experimenting with code. This self-directed learning eventually led him to transition into a research career.
Research at Google
In 2016, Ha joined Google as a research scientist. He was initially based in the United States, but he soon moved to Tokyo to help establish a local research presence. Over six years, he contributed to the growth of Google Brain in Tokyo, serving as a lead researcher and helping to build a team focused on deep learning and generative models.
During his time at Google, Ha collaborated with researchers including Llion Jones, who was a co-author of the seminal Transformer paper. Ha's own research spanned a range of topics, including neural architecture search, reinforcement learning, and world models. One of his notable projects was the development of a model that could generate images of digits and other simple objects, demonstrating the potential of neural networks to learn internal representations of visual data.
Ha also became known for his public-facing work. He wrote blog posts and gave talks that made complex ideas in generative AI accessible to a broader audience. His research on attention mechanisms, which are central to modern large language models, helped bridge the gap between academic theory and practical applications.
Move to Stability AI
In 2022, Ha left Google to join Stability AI, a London-based startup known for its work on open-source generative models. He was appointed head of research, where he oversaw the company's efforts to develop and release models like Stable Diffusion, which became widely used for text-to-image generation. Under his leadership, Stability AI expanded its research team and explored new applications of generative models.
However, Ha's tenure at Stability AI was relatively short. In June 2023, he resigned from the company, citing a desire to pursue new opportunities. His departure came amid broader organizational changes at Stability AI, which later faced financial and leadership challenges.
Founding Sakana AI
Shortly after leaving Stability AI, Ha co-founded Sakana AI with Llion Jones and Ren Ito. The company, based in Tokyo, was launched in August 2023 with the goal of developing AI systems inspired by natural intelligence. The name "Sakana" means "fish" in Japanese, reflecting the founders' interest in collective behavior and swarm intelligence.
Sakana AI focuses on creating models that can learn from smaller datasets and adapt to new tasks more efficiently than traditional approaches. The company has explored techniques such as evolutionary optimization and modular architectures, which allow different components of a model to be combined in novel ways. This approach has attracted attention from investors, and Sakana AI has raised significant funding to support its research.
As CEO, Ha has positioned Sakana AI as a leader in what he calls "nature-inspired intelligence." The company has released several open-source models and has collaborated with academic institutions in Japan. Ha has also been vocal about the importance of building AI that is aligned with human values, a topic he has discussed in interviews and public appearances.
Contributions to AI Research
Ha's research contributions span several areas of machine learning. He has published papers on neural architecture search, which involves automating the design of neural networks, and on world models, which are systems that learn to predict future states of an environment. His work on attention mechanisms has been influential in the development of Transformer-based models, which underpin many modern AI applications.
One of Ha's most cited papers, "World Models," co-authored with Jürgen Schmidhuber, introduced a framework for training agents in virtual environments using a compressed representation of the world. This work has been used in robotics and game-playing AI, and it has inspired further research in model-based reinforcement learning.
Ha has also contributed to the open-source community. He has released code and models under permissive licenses, allowing other researchers to build on his work. His blog, which he maintained during his time at Google, remains a popular resource for students and practitioners.
Recognition and Impact
In 2025, Ha was named one of the TIME 100 Most Influential People in AI, an annual list that recognizes individuals who have made significant contributions to the field. This honor reflects his dual role as a researcher and entrepreneur, as well as his influence in shaping the direction of AI development in Japan and globally.
Ha's work has been featured in major media outlets, and he is a frequent speaker at conferences such as NeurIPS and ICML. He has also been involved in policy discussions about AI regulation, advocating for a balanced approach that encourages innovation while addressing potential risks.
Personal Life and Legacy
Ha is known for his interdisciplinary background, combining expertise in engineering, finance, and computer science. He has spoken about the importance of curiosity and lifelong learning, and he encourages young researchers to explore fields outside their immediate area of study.
As of 2025, Ha continues to lead Sakana AI, which has grown to become one of the most prominent AI startups in Japan. His journey from finance to AI research and entrepreneurship serves as an example of how diverse experiences can contribute to innovation in technology.