Machine Learning Tech Lead - Post-Training for AI Characters

About the Role

At Starley, we’re creating Cotomo—a voice-based AI platform where users can form emotional bonds with anime and manga-inspired characters, including those voiced by real Japanese seiyuu in official collaborations.

Cotomo is already one of Japan’s largest platforms of its kind, with over 1 million downloads. As we prepare to expand globally, we’re looking for a Team Lead in Machine Learning to help shape the core intelligence behind our characters.

You’ll guide a small ML team and work across voice, character, and product to ensure Cotomo’s AI is expressive, consistent, and emotionally resonant.

If you love anime, manga, or games—and want to bring that passion into your work—this is your chance.

Responsibilities

• Lead post-training and alignment efforts for Cotomo’s LLM and voice models.
• Design systems that ensure personality, memory, and emotional depth.
• Collaborate across teams to turn research into production-ready features.
• Stay current with LLM/ASR/TTS research and apply advances meaningfully.
• Fluency in Japanese for daily communication

Key Qualifications

• Hands-on experience with LLM post-training (SFT, LoRA, DPO, etc.).
• Strong coding skills in Python and ML frameworks like PyTorch.
• Experience leading or mentoring in ML teams.
• Product-focused mindset—ship fast, learn fast.

Preferred Qualifications

• Deep interest in anime, manga, or games.
• Open-source or published work in ML/LLM/Voice.
• Background in persona modeling or interactive agents.
• Experience in fast-moving startups or R&D teams.

Work Environment

• Based in our Akasaka, Tokyo office
• Primarily in-office with some remote flexibility
• Visa sponsorship

Compensation & Benefits

• From ¥9,000,000 JPY/year, depending on experience
• Performance-based stock options