The Sakana AI Recursive Self-Improvement (RSI) Lab is a dedicated research group within Sakana AI, tasked with redesigning the AI development process itself with AI.
The Sakana AI RSI Lab is a dedicated research group within Sakana AI, tasked with redesigning the AI development process itself with AI. By transitioning from static, human-led R&D to autonomous, self-improving intelligence engines, we are turning constraints into our greatest compounding advantage. Working directly with our Co-Founder and CEO David Ha (@hardmaru), we are building the definitive architecture for the next frontier of AI.
Please read our announcement about this moonshot effort.
The establishment of our RSI Lab marks a serious commitment to engineering the next great leap in computational intelligence: transitioning from human-led R&D to autonomous intelligence engines capable of driving their own advancement.. Our research spans frontier language models, world models that learn the dynamics of the physical world from real-world interaction data, and the recursive self-improvement loop that binds them.
This direction builds on a long history of research into learned internal models of the world. As early as 1990, Jürgen Schmidhuber explored recurrent neural networks that learned predictive models of an environment and could be used by a controller for planning; decades later, our CEO David Ha and Schmidhuber’s World Models demonstrated agents that could even learn policies inside the “dreams” generated by such models. Today, the RSI Lab is pushing this idea toward frontier-scale systems that can model, reason about, and ultimately act in the physical world.
Artwork from Jürgen Schmidhuber’s World Models (1990-) papers.
Building on Japan’s long history of leadership in robotics, and bolstered by Japan’s strategic push for sovereign AI capabilities, we are scaling our research and engineering resources in Tokyo to pursue this mission. As part of this broader effort, Sakana AI has recently become a member of the AI Robot Association (AIRoA), joining Japan’s growing ecosystem for robot learning and physical AI.
Japan’s long history of robotics innovation, from WABOT-1 and AIBO to ASIMO and HRP-5P, helped lay the foundations for today’s emerging era of Physical AI.
We are looking for exceptional, highly driven individuals to join us to advance this vision and to build the engine of recursive discovery.
Who We Are Looking For
We are seeking a select group of distinguished Frontier Research Scientists and Advanced Core Engineers to join our core team.
You are a recognized innovator in the global AI/ML community. We are looking for distinguished individuals whose past work have already demonstrated a global impact and are widely acknowledged to have fundamentally altered the trajectory of the AI/ML ecosystem. This includes pioneers in self-improving systems, world models, robot learning, and physical AI.
You have likely produced contributions at top frontier labs or pioneered foundational frameworks, and you are entirely unsatisfied with the status quo. You have a proven track record at the frontier, but you are ready to break away from standard benchmarking and brute-force scaling to build the self-improving future. Candidates will be expected to be physically based in Tokyo, but we will provide assistance with the Visa application process if needed. We will be offering both 1-year residency and full-time positions.
If you are excited about the mission of the RSI Lab but are earlier in your career or do not yet match this profile, we also welcome applications for 4-month internships within the RSI Lab through this same application. For internship candidates, we place greater emphasis on technical ability, research potential, creativity, and exceptional drive than on an already established track record. When applying, please indicate whether you are interested in a full-time position or an internship.
Sakana AI’s RSI Research
What You Will Do
- Discover fundamental new laws of machine intelligence that bend the scaling curve.
- Build world models that serve as verifiable simulators for agentic reasoning and planning, bringing self-improving AI into the physical world.
- Apply open-ended evolutionary dynamics to high-stakes algorithmic domains, such as cybersecurity.
- Build the systems and infrastructure backbone of the RSI loop, optimizing high-dimensional search pipelines, managing massive distributed compute topologies, and productionizing automated code-generation stacks at an extreme engineering scale.
If you are a visionary builder ready to relocate to Japan and engineer the engine of recursive discovery, we invite you to apply.
Join the RSI Lab
Note on Applications
Please note that this specific role is highly selective. If you are passionate about our mission but your background does not perfectly align with this frontier profile, we strongly encourage you to explore our other open positions, including the research-focused Software Engineer (R&D) position on our careers page.