[ The Studio ]

[ Manifesto ]
The screen was a window. AI is a doorway. We build generative realities that learn, adapt and respond to every person inside them.
We build from the model up — diffusion, LLMs, NeRFs and agents are not bolt-ons, they are the foundation.
Every AI interaction is designed with empathy, not automation — avatars that feel, not bots that bark.
We are ML engineers who love spatial designers who love story. The deliverable is always real, always running.
[ Trajectory ]
Founded with a mission: make spatial worlds intelligent. First custom diffusion model for 3D generation.
First NeRF-to-VR pipeline. First LLM-driven avatar. First Awwwards for an AI-generated spatial world.
Opened Tokyo AI Lab. Released HALO OS, a neural spatial UI kit with 400k+ downloads.
Apple Vision Pro AI launch partner. 28 researchers, 3 cities, billions of parameters trained.
AURORA — our largest generative concert, built entirely by AI and performed by neural avatars.
[ FAQ ]
Both. Half the team are ML researchers training generative models; the other half are spatial designers and engineers who deploy those models inside AR/VR products. We are explicitly AI-first.
We fine-tune and train custom models for every long-form engagement — diffusion for 3D, LoRA-tuned LLMs for in-character avatars and small distilled models for on-device inference. We use foundation APIs only for prototyping.
The client. Every contract includes a clean handover — model weights, training data lineage, evaluation harness and an on-device runtime that the brand can keep running indefinitely.
28 people across Bengaluru, Mumbai and Hyderabad, including 11 dedicated ML researchers. We deliberately stay small to keep model quality and craft uncompromised.