Senior Director, AI Model Lifecycle

Crusoe
San Francisco, CA - US / Sunnyvale, CA - US2026-04-24OnSite

About the job

The Senior Director for the AI Model Lifecycle team will undertake a pivotal role in establishing both a dedicated team and a comprehensive managed platform to oversee the entire application development lifecycle, with a particular emphasis on utilizing Machine Learning models, including Large Language Models (LLMs).

Responsibilities

- Building a Team of Machine Learning Experts and being the Site leader for the Model Lifecycle Team.

- Manage fine-tuning systems for large foundation models (SFT, PEFT, LoRA, adapters), including multi-node orchestration, checkpointing, failure recovery, and cost-efficient scaling.

- Implement and maintain end-to-end training pipelines for Large Language Models.

- RFT and Reinforcement learning to the fine tuning and training sections

- Distillation and reinforcement learning pipelines (e.g., preference optimization, policy optimization, reward modeling).

- Dataset, model, and experiment management: versioning, lineage, evaluation, and reproducible fine-tuning at scale.

Qualifications

Minimum

- Advanced degree in Computer Science, Engineering, or a related field.

- 10+ years of industry experience leading and driving impactful projects in the AI Space

- Lead and mentor a team of engineers with exceptional interpersonal skills, working autonomously while proactively collaborating with stakeholders at all levels.

- Experience in Generative AI (Large Language Models, Multimodal).

- Hands-on experience training, fine-tuning, and aligning LLMs using Reinforcement Learning and Reinforcement Fine-Tuning (RFT) techniques.

Preferred

- PhD in Machine Learning, Computer Science, NLP, or a related field strongly preferred

- Research publications at NeurIPS, ICML, ICLR, ACL, EMNLP, or impactful preprints in the LLM post-training space

- Proficiency in Golang or Python for large-scale, production-level services and PyTorch

- Contributions to open-source AI projects such as vLLM or similar frameworks.

- Performance optimizations on GPU systems and inference frameworks.