Member of Technical Staff - Research Engineer, Frontier AI Robotics

Amazon•San Francisco, CA

3d•$150,000 - $300,000

About The Position

At Frontier AI & Robotics, we're building the future of intelligent robotics through frontier foundation models and end-to-end learned systems. We tackle some of the most challenging problems in AI and robotics, from developing sophisticated perception systems to creating adaptive manipulation strategies that work in complex, real-world scenarios. We leverage Amazon's computational infrastructure and rich real-world datasets to train and deploy state-of-the-art foundation models. Our work spans the full spectrum of robotics intelligence – from multimodal perception using images, videos, and sensor data, to sophisticated manipulation strategies that can handle diverse real-world scenarios. We're building systems that don't just work in the lab, but scale to meet the demands of Amazon's global operations. As a Senior Research Engineer embedded in our science team, you'll be instrumental in transforming innovative research into high-performance production systems. You'll collaborate directly with scientists to build and optimize large-scale transformer models for robotics applications. In this role, you'll balance deep technical optimization work with strategic input on model architecture decisions, ensuring our innovative robotics models are designed with performance in mind from the ground up. You'll leverage PyTorch and NVIDIA's acceleration stack and other compilation techniques to tackle ambitious performance targets, working at the intersection of large language models and real-world robotics applications.

Requirements

5+ years of non-internship professional software development experience, or Bachelor's degree in computer science or equivalent
5+ years of programming with at least one software programming language experience
5+ years of leading design or architecture (design patterns, reliability and scaling) of new and existing systems experience
Experience as a mentor, tech lead or leading an engineering team
Bachelor's degree in computer science or equivalent
Expertise in Python, PyTorch, and CUDA programming
Experience with TensorRT or similar ML optimization frameworks
Ability to optimize ML models for production
Experience working directly with research teams
Experience working on distributed training for the foundation model to make them stable, reliable, and performant

Nice To Haves

5+ years of full software development life cycle, including coding standards, code reviews, source control management, build processes, testing, and operations experience
Track record of improving research productivity through infrastructure design or process improvements.
Experience in ML infrastructure such as Pytorch, Megatron, TorchTitan, etc
Experience with ML compilers (ONNX Runtime, TVM, etc.)
Experience with transformer model optimization or RL framework development
Background in performance profiling and optimization
Track record of building robust monitoring systems
Experience with large-scale ML serving systems

Responsibilities

Design, implement, and optimize distributed training systems that scale across thousands of GPUs and nodes for large-scale training workloads.
Develop high-performance optimizations to maximize throughput and efficiency.
Develop reusable frameworks and libraries to improve training reproducibility, reliability, and scalability for new model architectures.
Establish standards for reliability, maintainability, and security, ensuring systems are robust under rapid iteration.
Collaborate with researchers to influence model architectures for optimal hardware utilization.
Develop comprehensive benchmarking frameworks to measure and optimize model performance.
Optimize transformer blocks using custom CUDA kernels and TensorRT optimization techniques.
Partner with scientists to analyze model architectures and propose efficiency improvements.
Implement and benchmark various optimization strategies for large-scale models.
Debug performance bottlenecks using NVIDIA profiling tools.
Participate in technical discussions about new model architectures with the science team.
Manage pre/post training runs and continue improve system stability and throughput.
Prototype new acceleration approaches using emerging compilation frameworks.