Member of Technical Staff – ML Ops

Member of Technical Staff – ML Ops

Requirements

• Bachelor’s or Master’s degree in Computer Science, Machine Learning, or related field
• Strong programming skills in Python and systems programming
• Experience with distributed systems and scalable infrastructure
• Track record of building reliable, performant large-scale ML systems
• Deep expertise in PyTorch internals and distributed training frameworks (FSDP, DeepSpeed)
• GPU cluster management and optimization
• Performance profiling and systems optimization
• CUDA programming and kernel optimization
• Containerization and orchestration (Docker, Kubernetes)
• ML model serving and deployment at scale
• Language models and attention mechanism optimization
• Video and audio processing pipelines
• Large-scale diffusion models
• Love diving deep into complex systems optimization challenges
• Take ownership of critical infrastructure while collaborating effectively
• Get excited about pushing the boundaries of ML system performance
• Want to work directly with researchers on cutting-edge ML problems
• Thrive in fast-paced, research-driven environments

Responsibilities

• Develop and optimize distributed training frameworks integrating multiple modalities (video, audio, text, and structured metadata)
• Build flexible systems for cross-modal training orchestration and efficient experimentation
• Design reproducible training environments with versioned dependencies and configurations
• Implement comprehensive testing frameworks for validating model training correctness and performance
• Create infrastructure for systematic model quality assessment and performance benchmarking
• Design and implement flexible training orchestration systems that balance research agility with large-scale model training
• Build robust monitoring and observability systems for complex training and inference pipelines
• Design and manage GPU clusters optimized for distributed training of multimodal models
• Build out comprehensive automated metrics collection and alerting across our ML stack
• Profile and optimize model training throughput using mixed precision, gradient checkpointing, and advanced memory techniques
• Develop custom CUDA and Triton kernels to accelerate critical compute paths
• Implement creative solutions for cost optimization across spot instances and reserved capacity
• Design and optimize real-time inference systems enabling fast research iteration cycles
• Build infrastructure enabling rapid testing of research hypotheses
• Create systems supporting close collaboration between infrastructure and research teams
• Develop frameworks for reproducible research experimentation
• Enable seamless deployment of research innovations to production

Preferred Qualifications

• Strong experience in some or all of the areas listed in the requirements