Pasadena, California, United States
Senior Machine Learning Engineer with 10+ years at Google, specializing in improving model performance through data quality analysis, bias correction, and rigorous evaluation design. Experienced in developing and fine-tuning deep learning models (TensorFlow, DNNs, BERT), designing evaluation pipelines, and building scalable data processes to ensure models generalize well across diverse populations. Proven ability to partner across product, legal, and engineering teams to align model improvements with business and regulatory needs. Strong advocate for interpretable, fair, and reliable ML systems, with a passion for identifying hidden weaknesses in data and designing pragmatic solutions that balance accuracy, fairness, and real-world usability.
Kids' Accounts * Improved safety related to data handling for underage users. * Worked to ensure compliance with government regulations for kids' data. * Fine-tuned LLMs to understand user-produced text signals. * Developed DNN multi-class and binary models in TensorFlow models to understand user demographics. * Used SQL, Python, and additional internal tools for model quality analysis, training/eval data analysis, and improving metric accuracy. Android Auto Team * Developed NLP models to label user intent for voice commands issued to cars. * Localized intent models in Portuguese, Norwegian, Spanish, and Russian, enabling the launch of Android Auto in these regions. Signals Team * Developed and improved models to understand content in text from a broad range of sources. Model formats included linear, FastXML, DNN, and Bert, with substantial work in TensorFlow. * Evaluated models using pre-labeled data as well as presenting model output to human raters for scoring. * Oversaw process of refreshing existing models to incorporate new data. * Streamlined and automated processes for frequent updates and improvements to existing models. * Identified and updated/deprecated/refactored legacy uses of team’s models to improve code health, ensure proper use, simplify updates, and ensure clients’ compliance with legal guidance.
Developed and supported utilities to assist medical facilities in understanding trends specific to their patients and take action based on this information. Developed user-facing as well as server code. Performed QA on changes created by other developers. Certified in EpicCare Ambulatory. Development experience in Reporting Workbench.
Processed and analyzed data at PredPol to provide an understanding of the effectiveness, usage, and potential for the software. Interpreted results to make them accessible to users. Explored new ideas and approaches to any mathematically-related questions and algorithms in order to advance understandings and more effectively tackle problems. Performed QA for sites, helping identify and correct any problematic or unusual behavior.
Produced schematics, box drawings, and network diagrams for automated equipment.