Inside Sakana AI’s Product Team: Autonomy, Tooling, and Hiring in Tokyo

Sakana AI, the Tokyo-based artificial intelligence research lab known for its evolutionary model merging and foundation models, has published a detailed look inside its Product Team. The blog post, based on interviews with four team members including Head of Product Sota Omura, offers a rare window into how the company’s research is turned into usable products, what tools the team relies on, and what it looks for in new hires.

The Product Team, which includes product managers, applied research engineers (AREs), software engineers (SWEs), designers, and sales staff, is responsible for bridging the gap between Sakana’s research output and real-world business applications. As of September 2026, the team is diverse in nationality and role, and operates with a strong emphasis on autonomy and on-site collaboration.

Key facts

How the team works: remote flexibility with on-site expectations

The Product Team encourages face-to-face communication but does not mandate a rigid schedule. Some members work nearly five days a week in the office, while others arrive between 11:00 and 12:00 or leave around 17:00. There are no fixed core hours—the standard day is 10:00 to 19:00—and the company expects individuals to manage their own productivity. However, because the team operates primarily on-site, remote workers are expected to close any information gaps on their own initiative.

Meetings are kept lean. A full Product Team all-hands occurs about once a month. Individual teams set their own cadences: some run short daily syncs, others hold weekly retrospectives. Separate regular meetings cover KPI reporting and progress updates with Omura, and there are sessions to discuss product direction directly with CEO David Ha.

Internal communication is bilingual. On teams with English speakers, English is encouraged for both documentation and spoken communication, but Japanese speakers are not required to use English with each other.

Tooling: Claude Code, Devin, and a culture of experimentation

The team’s tool stack reflects a strong preference for modern AI-assisted development. Slack, Claude Code, Codex, Figma, Confluence, Google Workspace, Jira, Devin, GitHub, Typeless, and Adobe Creative Cloud are all in active use. The Product Team actively encourages trying new products, and members can bring in tools they want to use.

This openness to experimentation aligns with the team’s broader culture. Many members enjoy building things outside of work—designing caps, T-shirts, and stickers featuring product logos—and the team has recently welcomed more members from outside Japan, increasing diversity. Planned additions include dedicated sales and marketing roles.

Autonomy and role boundaries: ARE vs. SWE

The question engineers ask most often, according to the blog, is about autonomy. The answer is clear: a lot. Broad direction is set at the company, team, and product levels, but within that framework, individuals have significant freedom. One ARE has taken on a product management role for the API Platform based on their own interests.

The boundary between AREs and SWEs is intentionally flexible. AREs build the core technical components of a product, while SWEs build the surrounding system—infrastructure, monitoring, QA, testing, data governance, security, and stakeholder communication. Because the team has relatively few SWEs for the number of products it runs, the environment suits full-stack generalists more than narrow specialists.

AREs on the Product Team differ from those on the Applied Team. Product Team AREs focus on generality when benchmarking and developing AI technology, working closely with the Research Team. Applied Team AREs tailor agent systems to specific customer requirements.

Hiring criteria: technical judgment and ownership

The blog details what Sakana looks for in its technical problem sets. For ARE candidates, the key is not simply solving the problem but understanding the strengths and weaknesses of each technique and explaining them in their own words. Product development involves daily decisions under constraints of accuracy, latency, cost, and UX, so quick and sound judgment is essential. Candidates are also evaluated on their ability to design appropriate evaluations and benchmarks.

For SWE candidates, the focus is on taking responsibility for decisions and actions. Using AI in development is now the norm, but Sakana looks for people who learn alongside AI rather than handing everything off. The technical assignment requires baseline quality, but the emphasis is on explaining technology and design choices in the candidate’s own words.

Head of Product Sota Omura outlined a broader set of attributes he hopes to see in team members: deep curiosity, strong communication skills, the ability to perform without supervision, self-directed learning grounded in metacognition, ownership that expands rather than narrows responsibility, the courage to take calculated risks, a spirit of fellowship, and a tendency to find good faith in others. His summary: “In short, someone who is kind.”

Why this matters for AI practitioners and job seekers

For AI engineers, product managers, and researchers considering roles at a Tokyo-based AI lab, this post provides concrete detail about the working environment, tooling, and expectations. Sakana AI’s approach—blending research and product, encouraging tool experimentation, and valuing autonomy—reflects a growing trend among AI labs that aim to move quickly from research to deployment. The explicit hiring criteria also offer a useful benchmark for anyone preparing applications for ARE or SWE positions in the AI industry.

Source: Sakana AI Blog — Inside Sakana AI’s Product Team (https://www.sakana.ai/inside-product-team/)