We are looking for a strong engineer to work on the infrastructure for LLM-based data labeling (crowd). LLM labeling is critical for the quality of Search – Yandex's main service.

Our team works on the semantic part of Search. We:

• Help Search better understand user queries, web documents, and their relationships
• Work at the intersection of NLP and ranking
• Help build Yandex's agent systems
• Use hundreds of GPUs for experiments
• Impact tens of millions of people with our releases

What tasks await you

Improve the efficiency and fault tolerance of the LLM infrastructure You will optimize model serving, fine-tuning pipelines, and crowd orchestration.

Develop observability and monitoring You will work on improving logging, alerting, and metrics for LLM tasks and labeling systems.

Develop and maintain ML infrastructure for model fine-tuning You will automate training cycles, collect and validate data.

Ensure the stability and reliability of LLM components in the data labeling infrastructure It is important to think through how to ensure maximum pipeline stability, considering potential system weak points.

More about ML at Yandex – in the channel Yandex for ML

We expect that you

• Can confidently perform system design for simple systems
• Have developed backends for ML services
• Understand LLM serving principles
• Have performed SRE tasks
• Write good, maintainable code in Python
• Can learn new things quickly
• Are ready to take ownership in your area of responsibility