Senior Software Engineer - MAIA System Infrastructure

MicrosoftUS

Full-time

Posted 3 days agoStill open, checked 1h ago

About the job

Own the PCIe data path and the host/device interface. Design and implement direct memory access (DMA) submission, completion, and error reporting across the host driver and device firmware. Specify what a completion guarantees and when a caller can safely reuse a buffer. Design memory and synchronization primitives around the chip's addressing, ordering, and visibility guarantees. Measure bandwidth, latency, and contention across its memory fabrics to guide decisions about mappings, queue depths, and transfer sizes. Own system-level chip integration and management. Bring up device initialization and firmware loading, expose device health to the host, and implement reset and recovery paths that preserve isolation and account for in-flight work. Build observability and developer tools across the host/device boundary. Give engineers the counters, traces, and crash dumps to distinguish a backed-up host queue from a stalled device and determine where progress stopped. Lead interface reviews and mentor engineers as new capabilities move from bring-up to production. Verify interface guarantees in simulation and on silicon with stress tests, fault injection, and performance baselines. Work with runtime and model engineers when a particular workload exposes a bottleneck in the platform. For example, test whether batching DMA submissions closes idle gaps on the PCIe link, then verify that the change helps the workload that exposed them. Bachelor's Degree in Computer Science or related technical field AND 4+ years technical engineering experience with coding in languages including, but not limited to, C, C++, C#, Java, JavaScript, or Python OR equivalent experience. Experience with PCIe, DMA, memory fabrics, memory-mapped I/O, or accelerator memory systems. Experience developing and shipping systems software in C, C++, or a comparable language for device drivers, firmware, operating systems, or high-performance I/O. Experience diagnosing a performance or correctness issue involving concurrency, memory behavior, or data movement, and validating a fix through tests or measurements. Experience designing or evolving a low-level interface with engineers responsible for adjacent hardware or software components. Experience bringing up and managing devices across firmware and Linux drivers, including initialization, health monitoring, fault isolation, or recovery. Experience building low-level debugging or profiling tools, or validating hardware/software interactions in simulation and on silicon. Familiarity with how accelerator runtimes use device interfaces, or experience working with runtime engineers on a workload-specific performance issue. Contributions to open-source systems software are welcome.