Software Engineer, GenAI Frameworks

Meta is building the infrastructure and systems that power machine learning at scale across its family of products, including Feed, Reels, Ads ranking, and generative AI services. The Systems ML Engineering team sits at the intersection of ML and systems software, designing and optimizing the training and inference pipelines, distributed execution frameworks, and data processing systems that enable researchers and product teams to iterate quickly and deploy reliably.

In this role, you will contribute to the full lifecycle of ML systems software — from designing scalable data pipelines and distributed training infrastructure to optimizing model serving latency and throughput — directly impacting the quality and speed of ML-powered experiences for billions of people.

Design and implement scalable systems for distributed ML training and inference, including data ingestion pipelines, feature processing, and model serving infrastructure Develop and optimize ML platform components such as training orchestration, gradient communication, and checkpoint management across large-scale distributed environments Profile and diagnose performance bottlenecks across the ML stack, including data loading, preprocessing, forward and backward passes, and serving latency Write automated tests covering expected behaviors, failure modes, and error paths for ML systems components, and build monitoring and alerting for production anomalies Collaborate with ML researchers and product engineers to translate model requirements into reliable, high-throughput system designs Own technical design for features and components within ML infrastructure, evaluating trade-offs between throughput, latency, cost, and engineering maintainability Participate in staged rollouts of ML system changes using feature flagging and A/B testing frameworks, monitoring key metrics and responding to regressions Contribute to code quality through code reviews, clear technical documentation, and consolidation of duplicative implementations across the ML systems codebase Support on-call rotations for owned ML infrastructure, investigating production incidents and contributing detailed retrospectives to prevent recurrence