Staff Software Engineer, ML Systems

We are looking for a Staff-level engineer to serve as the strategic bridge between our Machine Learning Product teams and our Core Infrastructure. In this role, you will own multiple Product teams’ ML Systems, covering Efficiency and Reliability, for our core machine learning pipelines.
You will act as a "force multiplier." While our ML researchers focus on novel model architectures and signal quality, you will focus on the ecosystem that enables them: optimizing training loops on our distributed cluster management system, streamlining data ingestion, and eliminating friction in the build and version control workflows. You will also write the high-performance "glue" that binds our model logic to our massive-scale proprietary infrastructure.

You will:

• Strategic Infra Roadmap: Partner with Core Infrastructure teams to influence the roadmap for compute, storage, and scheduling. Translate the ML team’s product requirements into concrete infrastructure requests (e.g., accelerator topology needs, storage throughput requirements) and prioritize them effectively.


• Unblocking & Reliability: Serve as the technical escalation point for production blockers. Troubleshoot complex failures that span the stack—from Python-level OOM errors in training jobs to underlying cluster scheduling, containerization, or network latency issues. Instrument modules to actively monitor and recover from instability..


• Efficiency & Performance: Profile and optimize end-to-end pipelines. Identify bottlenecks in data loading (e.g., fetching data from distributed storage to accelerators) and implement C++ optimizations where Python overhead is too high.


• Tooling & Automation: Build robust CLI tools and middleware to improve the "inner loop" of ML development. Automate tedious tasks in remote development environments to simplify change management and validation.


• Cross-Functional Integration: Write the necessary shims and wrappers to integrate new Core Infra features into the ML stack before they are officially supported, allowing the team to move faster than the platform baseline.

You have:

• Expert C++ & Python: Strong proficiency in C++ (system performance, concurrency, memory management) and Python (ML modeling, scripting). Ability to write production-ready code in a large-scale monorepo environment.


• Distributed Systems Fundamentals: Deep understanding of resource management (CPU/RAM/Accelerator isolation), job scheduling, RPC subsystems, and distributed storage.


• Debugging Expertise: Fearless approach to debugging "black box" system issues. Comfortable using profilers and digging into system logs to diagnose contention, deadlocks, or memory leaks.


• Leadership & Influence: Demonstrated ability to drive technical consensus across teams. Experience defining engineering standards, mentoring senior engineers, and managing complex stakeholder relationships.

We prefer:

• Experience with custom/proprietary build systems (e.g., Bazel-like environments).


• Experience optimizing large-scale ML workloads on custom AI accelerators (TPUs/GPUs).


• Familiarity with low-level serialization formats (e.g., Protocol Buffers).


• Background in optimizing remote development environments or interactive notebook workflows.