As Principal AI Software Architect for Microsoft MAIA accelerators within the SPARC team, you will play a pivotal role in enabling and optimizing machine learning model training workflows on our cutting-edge hardware. Your responsibilities will include bringing up and validating training processes on MAIA accelerators, collaborating with engineering and research teams to understand MAIA accelerator architecture, and enabling training recipes developed for Microsoft’s first-party accelerators. You will utilize your hands-on experience with PyTorch and at least one of Triton or CUDA to ensure efficient and scalable deployment of large models. This role is ideal for candidates who thrive in a dynamic environment and are passionate about advancing AI infrastructure and model performance on next-generation hardware platforms.

Leads by example across teams and mentors others to produce extensible, maintainable, well-tested, secure, and performant code used across products that adheres to design specifications.

Leads efforts to continuously improve code performance, testability, maintainability, effectiveness, and cost, while learning about and accounting for relevant trade-offs.

Identifies best practices and coding patterns (e.g., leveraging state-of-the-art generative artificial intelligence [GenAI], approaches to source code organization, naming conventions) and provides deep expertise in the coding and validation strategy.

Creates and applies metrics to drive code quality and stability, appropriate coding patterns, and best practices.

Identifies and anticipates blockers or unknowns during the development process, escalates them, communicates how they will impact timelines, and then leads efforts to identify and implement strategies and/or opportunities to address them.

Reviews product code and test code to ensure it meets team standards, contains the correct test coverage, and is appropriate for the product or solution area.

Brings insight to code reviews to help improve code quality, coaching and providing feedback to develop other engineers' skills.

Conducts code reviews in a timely fashion that helps accelerate the pace of development on the team. Considers diagnosability, reliability, testability, and maintainability when reviewing code, and understands when code is ready to be shared or delivered.

Applies and reviews for coding patterns, security risks, compliance issues, and best practices in code reviews, providing feedback on code to drive adherence to best practices.

Uses automated source code analysis tools that are incorporated into the build/development process.