Training Developer Specialist
Canon USA
AI Anime Researcher, LLM
Spellbrush
full-time
Posted on:
Origin: • 🇺🇸 United States • California
Visit company websiteJob Level
Mid-LevelSenior
About the role
- Join a small research team to define a new kind of video game experience
- Research and engineer LLMs that power anime characters and interactive narratives
- Collaborate with top researchers and engineers to design and build multi-modal models
- Implement and optimize training, evaluation, and post-training workflows (including roleplay-focused work)
- Design and analyze human feedback collection systems (RLHF-style or implicit signals)
- Build and maintain LLM infrastructure: data pipelines, checkpointing, sharding, memory optimizations, benchmarking, and checkpoint selection
- Optimize for real-time interaction and low latency when interacting with characters
- Contribute creatively to game design and character behavior; balance research exploration with product-focused development
- Prefer on-site collaboration in Tokyo (Akihabara) or San Francisco research offices
Requirements
- Published or industry experience with LLM post-training
- Deep understanding of LLM training and evaluation
- Bonus: post-training work focused on roleplay
- Bonus: experience with latency optimization for real-time interaction
- Experience designing or gathering human feedback (e.g., RLHF-style voting systems or implicit signal extraction)
- Experience with LLM infrastructure: data processing backend, deterministic/checkpointable data loaders, sharding, memory-saving techniques, scalable benchmarking, checkpoint selection
- Passion for anime and the anime aesthetic
- Comfortable working on small, fast-paced, in-person teams
- Willingness to relocate to San Francisco or Tokyo if necessary (visa sponsorship available)
