Five people in Shinjuku, Tokyo. We cover generative audio, character design, fan systems, music production, and the kind of obsessive attention to idol culture that makes the whole thing credible.
Spent years building real-time media infrastructure for the Japanese entertainment industry — close enough to the machine to see exactly who it kept out. Founded Paradigm AI in 2022 in Shinjuku with a specific thesis: fan elections, not talent scouts, should determine who performs.
Generative audio researcher and character systems engineer. Spent years studying voice synthesis and expressive animation before joining Taiyo to build the performance engine that turns a community vote into a real idol — voice model, visual rig, and stage presence included.
Built and ran fan communities at a music streaming platform where he watched voting mechanics get gamed, engagement systems feel hollow, and Discord servers fall apart. He joined Paradigm AI to build the community infrastructure that actually holds — election systems, moderation structure, and fan-first product decisions.
Designed characters for indie animation studios across Japan and South Korea before joining Paradigm AI. Every AIdeal member's visual identity — costume vocabulary, motion style, color palette — runs through Sora's direction. She works from community aesthetic input, not from a template.
Produced and arranged tracks for independent artists across Asia before Paradigm AI. Oversees the AI-assisted songwriting and arrangement pipeline — responsible for making each AIdeal single sound like it belongs to the member the community voted for, not like a generic vocal synth output.
We are a small team and every role is load-bearing. If you care about virtual idol culture, generative performance, or participatory fan systems — and you want to work on something that does not exist anywhere else yet — send us your CV.