Sakana AI Launches Business Development Division: Beginning Commercialization of AI Technologies

Sakana AI has spent the time since its 2023 founding chasing frontier research, the kind of work that gets cited more than it gets billed for. That changes now. The company has stood up a new Business Development Division whose entire job is to take that research and turn it into something clients actually pay for.

Running it is Hiroki Taniguchi, who previously served as Chief Data Officer at LINE Yahoo, where he pushed data and AI adoption inside a large organization. Under him sits a business team pulled from consulting firms, government agencies and corporate roles. Alongside that, Ryosuke Okuda, formerly CTO at Preferred Networks, leads an engineering and research team built from AI practitioners who've shipped things before. The new division starts at around 20 people. Combined with Sakana AI's existing research staff, that pushes total headcount to roughly 50.

Taniguchi's pitch is pointed at financial institutions specifically. Regulation there keeps getting more complicated, products keep multiplying, and yet a lot of the actual work is still done by people shuffling paper. He says he's confident Sakana AI's technology can meaningfully sharpen and speed up that kind of operation, and he's targeting this spring to start delivering services to clients. He also expects the division's headcount to grow two to three times its current size as things ramp up.

The technology bench they're drawing from is real. In August 2024 Sakana AI unveiled AI Scientist, a system meant to automate the research pipeline itself — generating ideas, running experiments, writing up papers, even handling peer review — collapsing work that used to take weeks. This January came TAID, a distillation technique for shrinking huge foundation models down to something practical. One product of that work, TinySwallow-1.5B, compresses a 32-billion-parameter model into a 1.5-billion-parameter version that's still sharp enough in Japanese to run quickly on a phone. There's also evolutionary model merging, which fuses multiple open-source models using a process modeled on biological evolution, surfacing capabilities that none of the individual models had on their own.

For now, the initial customers Sakana AI is chasing are financial institutions that have already invested in the company, where the plan is to apply AI agents against their toughest operational problems. The other early target is the public sector, defense included, where Taniguchi's team wants to take on work with real social weight rather than just easy wins.