Software Engineer, Agentic Model Serving, Robotics
$174k - $252k • Mountain View, CA, USA • Full-time
Posted 10d ago
About the job
Google's software engineers develop next-generation technologies that change how billions of users connect, explore, and interact with information. As a Software Engineer focusing on agentic robotics systems and model serving, you will bridge the gap between cutting-edge AI research and robust production systems by architecting and optimizing the full runtime stack—from on-robot harnesses and real-time control interfaces to edge and cloud serving infrastructure. You will build reliable, low-latency, and scalable solutions for real-world robotics applications, accelerating research initiatives and industry partnerships. You will thrive at the intersection of machine learning infrastructure, distributed systems, and robotics, managing complex end-to-end challenges from co-designing on-robot execution and bidirectional streaming pipelines to optimizing and scaling foundation model inference across edge and cloud accelerators.
Responsibilities
- Architect highly reliable and scalable full-stack model serving and deployment solutions spanning on-robot runtimes, local edge compute, and cloud backend systems, customizing for stateful bidirectional streaming, multi-rate agentic control loops, and fast model hot-swapping in real-world robotics deployments.
- Conduct end-to-end performance analysis, profiling, and optimization across the entire sensory-to-action loop, from on-robot sensor ingestion, transport, and real-time execution handoffs to KV-cache management, graph compilation, and accelerator inference for cutting-edge robotics foundation models.
- Partner closely with research scientists, hardware/control engineers, and organizational partner teams to co-design agentic harnesses, streaming APIs, and closed-loop deployment workflows that provide flexibility and reliability across various robotics applications.
Requirements
- Bachelor's degree or an advanced degree.
- 4 years of experience in software development.
- Experience working with Large Language Models (LLMs) or Vision-Language Models (VLMs) in a production environment.
- Experience optimizing model performance for inference on hardware accelerators (e.g., TPUs and GPUs).
- Experience with machine learning frameworks, particularly JAX or TensorFlow, on inference and serving.
- Experience developing and integrating Google’s production model serving solutions like Beyond and Uniserve.
- Experience integrating security best practices directly into ML infrastructure.
- Expertise in building, profiling, and optimizing complex, highly available, and secure software systems for end-to-end performance.
- Familiarity with robotics software stacks and demonstrated experience taking complex technical features from initial design through to production deployment.
- Strong, proven programming skills in Python and C++.
Benefits
- 15% bonus target
- equity
- benefits
