Spellbrush Tech Stack

AI-native game studio building generative tools for anime and tactics games

Computer Games San Francisco, California 11–50 employees Founded 2018 Privately Held

Spellbrush is a 11–50-person game studio in San Francisco founded in 2018, operating an AI-first technical stack built around PyTorch, JAX, and distributed inference (Kubernetes, Ray, Kafka, NATS across AWS/GCP/Azure). The company maintains niji・journey, a generative AI product, while developing in-house game titles anchored on turn-based tactics and storytelling. Core pain points—scaling inference for millions, GPU workload optimization, petabyte data ingestion—reveal a company stretched between consumer AI platform operations and game development, with infrastructure investment dominating the project roadmap.

Tech Stack 26 technologies

Core StackKubernetes Kafka Redis React TypeScript PyTorch Apache Spark Apache Airflow AWS Snowflake Redshift BigQuery Python NATS GPU WebGL WebGPU Unity JAX TPU Google DeepMind Ray GCP Azure Hive Bash

What Spellbrush Is Building

◆Challenges

Scaling inference for millions of users
Optimizing gpu workloads
End-to-end ml infrastructure
Distributed systems across multi-cloud clusters
New business expansion
Low-latency interaction
Scalable data processing backend
Petabyte data ingestion
Data pipeline scaling

▲Active Projects

In-house llm storytelling system
Ai companion for 3d adventure game
End-to-end ml infrastructure
Next-generation inference architecture
Game-ready models
Niji・journey updates
Character communication system for flagship game
Anime and video game tools
Single-page editors
Ai model integration

Hiring Activity

Minimal10 roles · 0 in 30d

Department

Engineering

Research

Data

Design

Ops

Seniority

Mid

Senior

Manager

Company intelligence

Find more companies like Spellbrush by tech stack, pain points and active projects

Get started free

About Spellbrush

Spellbrush develops games and generative AI tools for the anime and gaming communities. The studio is structured around three operating areas: game development (turn-based tactics titles with character communication systems), an in-house LLM storytelling engine, and the niji・journey platform (a generative AI tool for creators). The tech footprint—spanning PyTorch, JAX, TPU, BigQuery, and multi-cloud orchestration—signals heavy ML engineering work. The company hires across Japan and the United States, with research and engineering roles representing the majority of open positions.

HeadquartersSan Francisco, California

Company Size11–50 employees

Founded2018

Hiring MarketsJapan, United States

Frequently Asked Questions

What tech stack does Spellbrush use?

Spellbrush uses PyTorch and JAX for model training, Kubernetes and Ray for distributed compute, Kafka and NATS for event streaming, and Redis for caching. Data processing runs on Snowflake, BigQuery, and Redshift across AWS, GCP, and Azure.