ML Team Lead (AI Search) at AI Studio Yandex Cloud

We are looking for an experienced and proactive lead who will head the ML team and be responsible for developing core technologies for AI services of Yandex Cloud. The main focus of the team is creating multi-tenant search and classification models and service components, which are provided to cloud users as: * Standalone API (for example, Embeddings API, Classification API) * Integrated functions within OpenAI-like Responses and RealTime API (implementation of RAG scenarios, memory) * Technological foundation for end products (AI Guardrails, SpeechSense)

The role involves combining technical leadership and organizational management of a small distributed team, designing ML systems, as well as direct participation in development.

We offer: * Direct influence on the core technologies of the Yandex Cloud AI platform * High level of technical autonomy and freedom in tool selection (active use and development of open source) * The opportunity to build processes and culture in a new team practically from scratch * A strong R&D team and work in cross-functional v-teams on technically complex and interesting tasks

About the team

Subscribe to the Inside Yandex Cloud Telegram channel to learn more about our team and technologies!

What tasks await you

Team You will manage a team of ML developers: conduct performance reviews, 1-to-1 meetings, set goals, help developers with career development, grow the team's knowledge and experience. You will need to decompose product goals and turn them into a technical roadmap, formulate the team's quarterly plans and be responsible for their realism and stable execution. Develop and implement CI/CD for models, code review processes, testing, and release cycle of ML artifacts. Represent R&D teams within the v-team, ensure visibility and transparency of the R&D team's activities, form technical requirements and requests to related teams (for example, requirements for data and infrastructure).

Development You will be responsible for researching and developing SOTA models for RAG scenarios (search, ranking), classification (few-shot/zero-shot, guards) and memory, as well as their integration into inference infrastructure. You will bear direct responsibility for model quality metrics, as well as for the performance and stability of inference backends (including logging, monitoring, and code test coverage). In addition, you will personally participate in development: write code (up to 30–40% of the time), conduct code reviews, help with architecture and problem diagnosis, prepare releases of models and backends.

We expect that you

• Have managed an ML team for one year or more
• Have a deep understanding of modern ML (neural networks, transformers) with a focus on NLP, Information Retrieval, or Generative AI
• Have developed and brought ML services with high reliability and performance requirements into production
• Are proficient in Python and PyTorch
• Understand the full lifecycle of an ML model: from requirement gathering and data preparation to production operation
• Are proactive and able to independently form a technical backlog and roadmap based on product goals

Will be a plus

• Have used open source projects and contributed to them (especially to inference libraries such as TensorRT, TensorRT-LLM, vLLM, SGLang)
• Have experience with C++ and low-level optimizations (CUDA)
• Have worked in a distributed team
• Are able to or want to speak publicly, write technical articles, or maintain a blog to increase the external visibility of the team and product