Staff Machine Learning Engineer, Personalization

About the role

This role focuses on developing the intelligence behind Spotify's Home feed and Shortcuts features. You will lead technical strategy for recommendation systems and large language models to improve discovery for millions of listeners.

Key facts

What you'll do

• Manage and upgrade the machine learning systems and models that drive the Home feed and Shortcuts.
• Develop personalized recommendation features for new AI-driven user experiences.
• Execute the training, fine-tuning, evaluation, and optimization of large language models using SFT, distillation, and parameter-efficient methods.
• Collaborate with product managers, designers, and data scientists to plan and carry out experimentation strategies.
• Oversee A/B testing, system monitoring, and model performance to balance quality, reliability, and cost.
• Enhance data pipelines and ML platform infrastructure to support personalization at scale.
• Provide technical leadership in undefined areas and shape the long-term architecture of personalization systems.
• Mentor machine learning engineers to improve team output and technical standards.

Requirements

• 8+ years of professional experience in developing and deploying production-grade machine learning systems.
• Deep knowledge of ranking models, recommendation systems, and large-scale content discovery.
• Proficiency in Python and practical experience with PyTorch.
• Experience with LLM training, fine-tuning, and optimization techniques like SFT, distillation, and LoRA.
• Background in managing large-scale inference systems with a focus on latency, cost, and reliability.
• Ability to design, run, and analyze online A/B tests.
• Experience with distributed ML workloads using frameworks such as Ray, FSDP, or HSDP.
• Experience building data pipelines and orchestration workflows using tools like Flyte, Airflow, BigQuery, and cloud storage.
• Strong communication skills to influence technical decisions across different teams.

Skills & tools

• Python
• PyTorch
• Large Language Models (SFT, distillation, LoRA)
• Recommendation systems
• Distributed ML (Ray, FSDP, HSDP)
• Data orchestration (Flyte, Airflow, BigQuery)

Practical notes

• Spotify provides reasonable accommodations for the interview and application process upon request.
• The role requires collaboration within the Eastern Standard time zone.