Lead Machine Learning Engineer - Locals only

Lead Machine Learning Engineer - Locals onlySan Francisco, CA - HybridFull TimeOur core engineering team is looking for a hands-on ML Engineering Lead who thrives in early-stage, ambiguous environments. You’ve led ML systems from v1 to scale, and enjoy defining both the technical direction and the systems that power them.Your mission is toOwn and lead the ML/AI function end-to-end, setting technical direction and standards across the companyArchitect and guide the development of multi-modal, agentic AI systems powering real-world workflowsDefine and oversee evaluation frameworks, datasets, and performance metrics to continuously improve agent qualityDrive product ionization of ML systems, ensuring reliability, scalability, and compliance in real-world environmentsBuild and mentor a high-performing ML team over time, setting best practices across modeling, experimentation, and deploymentWho You Are8+ years of experience in ML/AI engineering, including time as a technical lead or managerProven track record of leading ML initiatives end-to-end, from problem definition → production deploymentDeep experience with LLMs and/or agentic systems, ideally in real-world, customer-facing applicationsStrong understanding of ML fundamentals (deep learning, transformers, model evaluation, tradeoffs)Experience scaling ML systems in production, including monitoring, iteration, and reliabilityDemonstrated ability to lead engineers, influence architecture decisions, and drive technical directionComfortable operating in early-stage, ambiguous environments with high ownershipStrong communication skills with the ability to translate complex ML concepts into clear decisionsBonus Points If YouHave experience building agentic systems, orchestration layers, or long-context reasoning systemsAre comfortable across the stack (data → modelling → infra → APIs)Have worked with both open-source and closed LLMs, including fine-tuning or retrieval systems (RAG)Have a strong product mindset and care deeply about real-world impact, not just model performance