Bill Jia

I provide sanitation and janitorial services at Google. And more specifically, I help to clear all issues to make Artificial Intelligence and Machine Learning great at Google and for the world. At Google, Core ML/AI partners with Google DeepMind very closely to design Gemini models and GenMedia models (e.g. Veo 3, Imagen 4), pre-train them, fine-tune them with various optimization algorithms, integrate them into various Google products and cloud customers' products, and then run all the ML inference workloads. Core ML/AI is also in charge of providing all software related to TPUs from low level stack (libTPU, XLA, JAX, PyTorch, XProf/XMon) to upper level pre-training/post-training/inference software for both Google internal use cases and external use cases. We run the largest training scale on the planet with hundreds of thousand ML accelerators to run single pre-training jobs. This enables us to product the world most advanced LLM models. We developed consolidated fine-tuning and inference platforms to steamline all post training jobs and ML serving traffic on unified platforms. Further, we own Google's foundational ML software stacks including ML Framework (JAX) and ML compiler (XLA) to lower LLM model codes onto various hardware with optimal performance. We also build ML agentic framework and develop a unifed service to all all Google and non-Google developers to build AI agents and ochestrate these agents together to serve customers' business needs.

- Develop and build Meta AI and machine learning framework, distributed learning/training, explore the cutting-edge machine learning training platform and prediction platform. - Develop and build Meta data infrastructure including data logging, ingestion, streaming, warehouse, data compute (spark and presto), and data monitoring and visualization services. - Managing Meta product performance and capacity to cope up with the hyper-growing user base/traffic and the fast product launch cycle. - Managing company-wide software/service/product efficiency efforts to reduce infra cost while supporting infrastructure expansion. - Planning Meta short term and long term data center, network and hardware capacity growth and expansion plan. - Define, engineer and validate the major hardware (OCP) introduced into Meta, running at scale in production.

- Develop and apply algorithms, statistical/mathematical optimization models to enhance the architecture, capacity, and performance of the system to support critical decision-making processes; - Develop machine learning algorithms to predict user behaviors/preferences to design the web portal to improve user engagements; - Provide in-depth online data analysis (patterning, trending, clustering, matching) for various Microsoft Windows Live services with one of the largest web user bases in the world (1 billion plus); - Decision support and analysis for user traffic and storage pattern, architecture design, network configuration, web user allocation, capacity management, storage management, etc; - Monitor the critical web services in the largest data center in the world runing 24/7; - Recommend DC expansion and server capacity purchasing plan with analyzed user traffic and storage growth pattern. - Three keynote conference/forum speeches and three patents filed at this position;