Staff Software Engineering, GenAI Applications, YouTube

Job description

Design and implement dynamic context construction strategies for RAG (Retrieval Augmented Generation) systems. Solve challenges related to limited context versus massive schema documentation, optimizing for precision, recall, and cost. Lead the evaluation and fine-tuning of embedding strategy (dense, sparse, and hybrid) to capture domain-specific YouTube terminology. You will decide how we represent data tables, column definitions, and debug logs in vector space. Build data pipelines that keep our semantic index fresh in real-time as YouTube’s data schemas evolve. Mentor Senior Engineers, drive technical roadmap planning, and collaborate with partners in Google DeepMind and YouTube infrastructure to adopt research. Minimum Qualifications: Bachelor's degree or equivalent practical experience. 8 years of experience in software development. 5 years of experience testing, and launching software products, and 3 years of experience with software design and architecture. 5 years of experience with machine learning algorithms and tools (e.g. TensorFlow), artificial intelligence, deep learning, or natural language processing. Preferred Qualifications: 8 years of experience with data structures and algorithms. Experience leading technical strategy for large-scale projects and mentoring executive engineers. Experience designing and implementing complex context retrieval systems (context scaffolding) for large language models in production. Proficiency in SQL. Understanding of vector space models, semantic search, and experience evaluating embedding techniques (e.g., dense versus sparse retrieval, re-ranking strategies). Ability to independently query massive datasets to diagnose model behavior, identify edge cases, and drive data-informed architectural decisions.