Tech Lead Manager ML Optimization

Job description

Job responsibilities

• Technical Leadership: Proactively study the SOTA model architectures and optimizations from the community and Google, for World Models, Diffusion + flow matching techniques, and translate them into measurable technical…
• Performance Analysis: Dev tooling innovation for model performance inspector in highly distributed training/inference setups, apply roofline analysis, understand the efficiency headrooms and drive work groups to…
• Strong Execution: Innovate high performance optimizations and tools for various models and large-scale training/inference including on future next-gen TPUs and low-bit precision training/inference setup, and ensure all…
• Cross-Team Leadership: Guide efforts across multiple teams and organizations to ensure seamless integration of data generation, model development, and deployment pipelines.
• Mentorship & Management: Act as a mentor to junior engineers, helping to grow their technical expertise and foster a culture of collaboration and engineering excellence.

Job requirements

• 10+ years of professional software engineering experience, with at least 5 years in machine learning infrastructure such as developing, training, deploying, and optimizing large-scale machine learning systems.
• Experienced using ML accelerator profiling tools to uncover performance bottlenecks.
• Solid experience in the development and optimization of machine learning infrastructure tools like DeepSpeed, PyTorch, TensorFlow, JAX, or similar frameworks.
• Deep understanding of state-of-the-art machine learning models and architectures such as autoregressive and diffusion transformers and familiarity with custom-kernels for diverse h/w compute based efficiency.
• Strong leadership skills with experience navigating cross-functional teams and providing technical leadership projects across multiple organizations.
• Excellent communication skills, both verbal and written, with the ability to translate complex technical concepts for a broad audience.
• A Master’s or PhD in Computer Science, Engineering, or a related field is preferred.