About the job
We are an applied research team focused on advancing efficiency across the AI stack, spanning models, ML frameworks, cloud infrastructure, and hardware. We drive mid- and long-term product innovation through close collaboration with research and product teams across the company. We communicate our research both internally and externally through internal technical reports, academic conference publications, open-source releases, and patents. Beyond producing research, we take responsibility for driving ideas through prototyping, validation, and production, with a bias toward real-world impact. This Principal Researcher will work across the full stack—from large-scale serving systems to hardware- and kernel-level optimizations—exploring algorithmic, systems, and hardware/software co-design techniques. Areas of focus include batching, routing, scheduling, caching, endpoint configuration, and GPU architecture–aware optimizations. This role emphasizes end-to-end ownership, with responsibility for identifying high-impact problems and driving research ideas through prototyping, validation, and deployment to deliver measurable customer impact.
Responsibilities
Formulate, develop, and evaluate new algorithmic and system-level approaches for end-to-end AI serving, using analytical modeling and large-scale measurement to study token-level latency, tail latency (p95/p99), throughput-per-dollar, cold-start behavior, warm pool strategies, and capacity planning under multi-tenant SLOs and variable sequence lengths.
Design and experimentally evaluate endpoint configuration and execution policies, including batching, routing, and scheduling strategies, tensor and pipeline parallelism, quantization and precision profiles, speculative decoding, and chunked or streaming generation, and drive the most promising approaches through robust rollout and validation into production.
Perform hardware- and kernel-aware optimization by collaborating closely with model, kernel, compiler, and hardware teams to align serving algorithms with attention/KV innovations and accelerator capabilities.
Build and benchmark experimental prototypes and large-scale measurements to validate research ideas and drive them toward production readiness; produce clear technical documentation, design reviews, and operational playbooks.
Publish research results, file patents, and, where appropriate, contribute to open-source systems and serving frameworks.
Qualifications
Minimum
Doctorate in relevant field AND 6+ years related research experience OR Master's Degree in relevant field AND 7+ years related research experience OR Bachelor's Degree in relevant field AND 9+ years related research experience OR equivalent experience. Other Requirements: Ability to meet Microsoft, customer and/or government security screening requirements are required for this role. These requirements include but are not limited to the following specialized security screenings: Microsoft Cloud Background Check: This position will be required to pass the Microsoft Cloud background check upon hire/transfer and every two years thereafter.
Preferred
Doctorate in relevant field AND 8+ years related research experience OR Master's Degree in relevant field AND 12+ years related research experience OR Bachelor's Degree in relevant field AND 15+ years related research experience OR equivalent experience. Experience publishing academic papers as a lead author or essential contributor. Experience participating in a top conference in relevant research domain. Demonstrated experience in designing and optimizing efficient inference systems, combining foundations in algorithmic optimization, parallel computing, and request orchestration under strict SLO constraints with deep knowledge of attention and KV-cache optimizations, batching and scheduling strategies, and cost-aware deployment. 3+ years of experience with machine learning frameworks (e.g., PyTorch, TensorFlow) and inference serving frameworks (e.g., vLLM, Triton Inference Server, TensorRT-LLM, ONNX Runtime, Ray Serve, DeepSpeed-MII). 3+ years of experience in GPU programming and optimization, with expert knowledge of CUDA, ROCm, Triton, PTX, CUTLASS, or similar GPU programming frameworks. Experience in C++ and Python for high-performance systems, with code quality and profiling/debugging skills. Research impact through publications and/or patents, coupled with hands-on experience taking research ideas through execution and delivery in production.