OSHIMOS:Living Charactersfor Everyone

We believe the next great storyisn't written for you.It's lived with you — one message at a time.

WHO WE ARE

OSHIMOS is an AI-native character lab. We build the models, the memory and the craft that let a character hold a conversation worth coming back to — then ship it to millions of people, in their own language, on their own terms.

Registered in Singapore. Building for the world.

Capabilities

Six systems we own end to end. No middleware between us and the experience.

Character Engine

Persona modelling that holds a consistent voice, motive and boundary across thousands of turns — not just the first ten.

Long-Horizon Memory

A character remembers what mattered and lets go of what did not. Continuity is a retrieval problem, and we treat it like one.

Model Routing

A multi-model gateway that picks the right model for each turn, balancing latency, cost and craft in real time.

Multimodal Generation

Portraits, scenes and voice generated in-context, so the character you imagine is the character you see and hear.

Evaluation & Experiments

Offline evals before release, online experiments after. Every change to a character is measured against the conversation it produces.

Safety & Trust

Layered moderation, age assurance and region-aware policy, built into the pipeline rather than bolted on at the edge.

OUR PRODUCT

What we ship

Matchat

AI Character Chat & Roleplay

Thousands of characters, written by a community and brought to life by our engine. From personality to conversation style, anyone can create an AI that feels truly alive.

RomanceFantasyMafiaSlow BurnEnemy to LoversOffice

Visit Matchat

HOW WE WORK

Craft over scale

Small teams, whole problems

One team owns a surface from model to pixel. No handoffs, no translation loss.

Ship weekly, measure always

Every release is an experiment with a hypothesis attached and a number that settles it.

Written in the open

Architecture decisions, evals and post-mortems are documents anyone inside can read.

CAREERS

Build the characters people come back to

We are hiring across model research, inference infrastructure, product engineering and trust & safety. Remote-friendly, Singapore-registered, English and Mandarin both work.

See open roles

Say hello

Partnerships, press, or a role you think should exist — reach us directly.

[email protected]