Develop and implement the company's AI computing strategy to ensure alignment with the company's overall business goals.Identify new market opportunities and technology trends to promote the development and application of AI technology.Oversee the construction and optimization of AI computing platforms, including the selection of hardware and software infrastructure.Ensure that the computing frameworks and tools used meet industry standards and are efficient, secure, and scalable… Show more Develop and implement the company's AI computing strategy to ensure alignment with the company's overall business goals.Identify new market opportunities and technology trends to promote the development and application of AI technology.Oversee the construction and optimization of AI computing platforms, including the selection of hardware and software infrastructure.Ensure that the computing frameworks and tools used meet industry standards and are efficient, secure, and scalable. Lead the research and development of AI-related technologies and promote the implementation of cutting-edge technologies in products and solutions.Collaborate with other departments (such as product management, data science, engineering, etc.) to promote cross-functional innovation projects.Pay attention to ethical, regulatory, and compliance issues in the field of AI to ensure that the company's AI development and application comply with relevant legal provisions.Promote the docking of the company's AI technology with industry standards and enhance the company's industry influence. Show less

In 2015, I was part of the early founding team that architected Microsoft's first gen AI silicon, the Maia 100 accelerator. As part of the journey, I recruitedand grew a team from just a few engineers to a vertical AI team, including~75 hardware architects, HW engineers, software developers, and AI/ML datascientists. This work produced strategic IP for narrow- and mixed-precisiontraining, strategies for parallelization that leverage data, model, pipeline, andPage 1 of… Show more In 2015, I was part of the early founding team that architected Microsoft's first gen AI silicon, the Maia 100 accelerator. As part of the journey, I recruitedand grew a team from just a few engineers to a vertical AI team, including~75 hardware architects, HW engineers, software developers, and AI/ML datascientists. This work produced strategic IP for narrow- and mixed-precisiontraining, strategies for parallelization that leverage data, model, pipeline, andPage 1 of 3expert parallelisms, software collectives, optimized runtime/kernels, tools forworkload and performance analysis, data types, etc.In 2017, I led the pathfinding organization for our AI supercomputing work inAzure. My role was to guide the end-to-end AI supercomputer infrastructureroadmap to meet the computing needs of our largest customers and internalMicrosoft teams. We collaborated regularly with customer leads, solutionproviders, to understand workload requirements across the full stack, anddevelop solutions based on the needs of hyper-scale generative pretrainingand inferencing. Show less

I joined as one of the early members of the Catapult team that pioneeredthe use of FPGAs as datacenter accelerators. I contributed to all aspectsof pathfinding and product development from requirements gathering,architecture definition to execution across the entire stack (RTL, PD, hostruntime, device kernels, drivers, RAS, telemetry, DRI, etc). Catapult wasshipped on over >1M Microsoft servers, and also later influenced Intel’s$16.7B acquisition of Altera.

I was recruited to Intel Corporation’s early computer architecture group led by ChuckThacker. I helped Chuck design and prototype AN3, a low-cost circuitswitched datacenter network. I also developed the concept for LINQits,a domain-specific accelerator. On real HW, I demonstrated that LINQitsachieves an order of magnitude perf/W gain over CPUs.