[ The Studio ]

WE TRAIN
AI
THAT YOU
CAN WALK IN.

Person wearing iridescent VR headset

[ Manifesto ]

The screen was a window. AI is a doorway. We build generative realities that learn, adapt and respond to every person inside them.

01

Neural-native

We build from the model up — diffusion, LLMs, NeRFs and agents are not bolt-ons, they are the foundation.

02

Human-first AI

Every AI interaction is designed with empathy, not automation — avatars that feel, not bots that bark.

03

Shippable intelligence

We are ML engineers who love spatial designers who love story. The deliverable is always real, always running.

[ Trajectory ]

From garage to galaxy.

  1. 2021

    Founded with a mission: make spatial worlds intelligent. First custom diffusion model for 3D generation.

  2. 2023

    First NeRF-to-VR pipeline. First LLM-driven avatar. First Awwwards for an AI-generated spatial world.

  3. 2024

    Opened Tokyo AI Lab. Released HALO OS, a neural spatial UI kit with 400k+ downloads.

  4. 2025

    Apple Vision Pro AI launch partner. 28 researchers, 3 cities, billions of parameters trained.

  5. 2026

    AURORA — our largest generative concert, built entirely by AI and performed by neural avatars.

[ FAQ ]

About the studio.

Is Metaverze an AR/VR agency or an AI lab?

+

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.

Do you train custom models or use off-the-shelf APIs?

+

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.

Who owns the AI models you build for a client?

+

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.

How big is the studio?

+

28 people across Bengaluru, Mumbai and Hyderabad, including 11 dedicated ML researchers. We deliberately stay small to keep model quality and craft uncompromised.