Principal Machine Learning Engineer, Applied AI

Join Lila Sciences in Cambridge, MA onsite, applying advanced ML to bring frontier AI into dependable, evaluation-ready workflows for customer scientific use.

Responsibilities

• Use deep, hands-on expertise to close the last-mile gap between model capabilities and customer-specific scientific workflows.
• Lead model post-training work, including SFT and RL approaches such as DPO, PPO, and GRPO to align outputs with customer requirements and feedback.
• Build and operate evaluation loops that assess quality, reliability, and customer fit, using patterns refined across prior projects.
• Translate customer learnings, data signals, and evaluation findings into repeatable model improvement cycles.
• Collaborate with AI researchers to convert model advances into reliable, usable capabilities.
• Work with Software to integrate model behavior into end-to-end product workflows.
• Debug complex model failures using traces, evaluations, customer context, and scientific feedback.
• Mentor engineers and share reusable patterns for model adaptation, evaluation, and deployment, based on experience shipping ML systems.

Requirements

• 2-3+ years of hands-on post-training experience, including SFT and RL methods such as DPO/PPO/GRPO, plus evaluation system design from repeated delivery.
• Strong software engineering fundamentals in Python and modern ML frameworks like PyTorch.
• Proven ability to debug ambiguous, high-stakes model behavior quickly using data, traces, logs, and qualitative feedback.
• Experience leading technical work across research and engineering teams.
• Deep familiarity with large language models, multi-modal models, or agentic AI systems.
• Clear communication for translating customer needs into technical plans and explaining complex model behavior to technical and non-technical audiences.

Technologies

• Python
• PyTorch
• SFT
• RL
• DPO
• PPO
• GRPO
• RLHF
• MoE

Bonus Points

• Experience adapting models for customer-facing or production workflows, especially in scientific, technical, or data-intensive domains.
• Experience with RL post-training such as RLHF, GRPO, or tool-augmented RL.
• Experience building evaluation harnesses, model monitoring, or quality dashboards.
• Experience training MoE architectures.
• Mentoring experience and a reputation as a go-to technical expert (not centered on people management or strategy ownership).

Location: Cambridge, MA (onsite)
Compensation: USD 252,000 - 336,000 per year