Yarong Mu

CTO

Kirkland, Washington, United States25 yrs 5 mos experience
Most Likely To SwitchHighly Stable

Key Highlights

  • Expert in privacy-preserving AI/ML algorithms.
  • Led successful projects in high-stakes environments.
  • Proven ability to influence cross-functional teams.
Stackforce AI infers this person is a leader in privacy technology and AI/ML solutions within the AdTech and SaaS industries.

Contact

Skills

Core Skills

Differential PrivacyFederated LearningData Monetization

Other Skills

Differentially Private Training Algorithm Creation & DevelopmentTensorFlowPyTorchDecentralized Applications (DApps)Auction DynamicsPricing StrategyBlue Ocean StrategyData GovernancePrivate Information RetrievalTrusted Execution EnclavesDevise incentive mechanisms & design accompanying dataConcurrency and programming modelsAdvertisingDisplay AdvertisingMobile Advertising

About

Leadership & Collaboration - Strong communication, collaboration, and organizational skills to effectively work with senior leadership. - Proven ability to influence and build consensus across diverse teams, including research scientists, MLEs, SWEs, PMs, and cross-functional teams (BD, legal, PR, marketing, UXR). - Experience in identifying and growing key stakeholders, managers, and team members. Strategic Project Management - Expertise in independently managing and leading complex, multi-year privacy-preserving projects involving multiple internal and external stakeholders with varying interests. - Demonstrated problem-solving skills to tackle complex system challenges and drive long-term strategic initiatives. System & Privacy Expertise - In-depth knowledge of large system analysis, including the impact of changes in interconnected systems handling super high QPS throughput. - Extensive experience in privacy technologies such as Differential Privacy, Federated Learning, Federated Analytics, Private Information Retrieval, Trusted Execution Environments, and Secure Multi-Party Computation. - Proficient in DP budgeting, DP notions, DP techniques, DP ops, and DP cost functions, with hands-on experience in TensorFlow, PyTorch, JAX on various architectures (CPUs, GPUs, confidential CPUs like SEV-SNP/TDX, and confidential GPUs like Nvidia H100). Engineering & Cloud-Native Architecture - Solid understanding of cloud-native best practices, including control-flow/data-flow separation, storage/processing separation, micro services, server-less infrastructures, and sandboxing. - Expertise in business logic isolation design, ensuring verifiable privacy and security for compliance/audits while maintaining proprietary business logic. Team Management & Leadership - Proven track record in managing, developing, and growing engineering teams. - Demonstrates a leadership style that leads by example and by providing context.

Experience

25 yrs 5 mos
Total Experience
5 yrs 7 mos
Average Tenure
13 yrs 2 mos
Current Experience

Google

Senior Staff Software TLM

Mar 2013Present · 13 yrs 2 mos · On-site

  • Own privacy-preserving AI/ML algorithms and platform development within Google’s Privacy Sandbox product area, including both Differentially Private (DP) training and inference.
  • Oversaw the design, development, and deployment of decentralized (Federated Learning) and centralized (Trusted Execution Environments) algorithms, supporting both offline (batch) and online (incremental) training functionalities for internal and external customers.
  • Key responsibilities include leading both DP recommender systems and DP LLM projects.
  • Advanced the state of the art in Differentially Private Stochastic Gradient Descent (DP-SGD), improving utility by closing performance gaps between private models and non-private counterparts.
  • Conceived multiple novel, industry-leading DP training algorithms that outperform DP-SGD in utility and cost, making them cheaper to train and serve.
  • Founded and landed OnDevicePersonalization (ODP), a decentralized DP federated training and serving platform that powers Android 13+, enhancing consumer privacy by keeping user data on devices while providing personalized experiences.
  • In the past
  • Led to the successful, on time, and stable completion of projects including DV360 (Google’s Demand Side Platform), GSN (Google Search Network), GDA(Google Display Network), and AM360 (Google’s Supply Side Platform).
  • Scoped and developed ad systems in targeting, rendering, auctions, trafficking, reporting (billing, auditing, verification), and optimization (prediction, bidding, pricing).
  • Strategically positioned DV360 as the leading Demand Side Platform for efficient buying of premium video inventories by winning over major non-Google Supply Side Platforms resulting in them converting their premium inventories into formats that DV360 recommends.
  • Continuously improved DV360 auction efficiency by designing multiple new auction algorithms which significantly increased winning rates, frequently by more than an order of magnitude.
Differentially Private Training Algorithm Creation & DevelopmentFederated LearningDifferential PrivacyTensorFlowPyTorchDecentralized Applications (DApps)

Bloomberg l.p.

Senior Software Developer

Aug 2010Mar 2013 · 2 yrs 7 mos · Trading Solution OPEN/Gateway

General atomics

Research Collaborator

Feb 2007Nov 2009 · 2 yrs 9 mos · San Diego, CA

  • Research Collaborator at General Atomics, the National Fusion Facility
  • Worked on configuration, calibration, improvement, data acquisition, analysis and interpretation of outputs of a variety of edge diagnostics on DIII-D:
  • Cameras, such as the DiMES TV Cameras, Filterscopes
  • Spectrometers, such as the Multi-Chord Diverter Spectrometer
  • ASDEX Gauges
  • Participated in and familiar with DIII-D daily operation routine
  • Wrote customized code for each piece of edge diagnostics for data output and analysis
  • Cooperated with GA, Oakridge and Livermore national laboratories in designing and conducting edge fusion experiments:
  • The DIII-D DiMES Porous Plug Injector MkII Experiment (2007)
  • The DIII-D thermal oxidation experiment

University of toronto

PhD

Jan 2004Nov 2009 · 5 yrs 10 mos · Toronto, Ontario, Canada

  • University of Toronto Institute for Aerospace Studies (GPA 4.0)
  • Doctor of Philosophy in Plasma Physics and Nuclear Fusion
  • Thesis work focused on 3D-DIVIMP, a Physics Software Engineering project intended to model Impurity Particle Production, Evolution, Transportation, and Deposition
  • A complete application (100,000 lines, Fortran, C, MATLAB) was constructed with all facets of software construction present including Structural Design, Implementation, Testing, and Documentation
  • Generates graphical plots to ease human analysis of results
  • Designed and Implemented from scratch resulting in a sizeable and mature code-base
  • Uses various online database queries to determine impurity evolution
  • Highly configurable to allow use of a wide library of physical models
  • Includes diagnostic simulation for DIII-D edge diagnostics, such as MDS and DiMES TV
  • Thesis work also focused on the 2007 DIII-D DiMES Porous Plug Injector MkII experimental data analysis and simulation; as well as the 2010 DIII-D DiMES Lithium experimental data analysis and simulation
  • Was awarded the Connaught Scholarship

University of science and technology of china

Bachelor of Science

Sep 1999Jul 2003 · 3 yrs 10 mos · Hefei, Anhui, China

  • University of Science & Technology of China (GPA 3.82, Major: 3.90)
  • Bachelor of Science in Physics from the Department of Modern Physics
  • Built a Superconducting Diode Laser for a 677nm wavelength for lab plasma study
  • Coursework: Analytic Geometry & Algebra, Function Theory of Complex Variables, Digital Signal Processing in Physics, Computational Physics, Theoretical Mechanics, Electrodynamics, Probability Theory and Math Statistics, Advanced Quantum Mechanics, Thermodynamics and Statistical Physics, Atomic Physics, Plasma Physics, Nuclear & Particle Physics, Optics, Solid-state Physics, Radiation Measurements, Methods of Nuclear Electronics

Education

University of Toronto

PhD — Plasma Physics and Nuclear Fusion

Jan 2004Jan 2009

University of Science and Technology of China

BS — Modern Physics

Jan 1999Jan 2003

Stackforce found 100+ more professionals with Differential Privacy & Federated Learning

Explore similar profiles based on matching skills and experience