Open to opportunities

Yumi User

@yumiuser2

Message

Senior machine learning engineer specializing in applied ML, NLP, computer vision, and production systems.

United States

Message

What I'm looking for

I seek roles building scalable, production ML systems (NLP/CV/ads) with cross-functional teams, focusing on impactful products, rigorous evaluation, and deployment to cloud or edge.

I am a Senior Machine Learning Engineer with over nine years of experience building and shipping production ML systems across ads, NLP, and computer vision domains. I combine a Ph.D. in Statistics with hands-on expertise in large-scale prediction systems, transformer-based NLP, edge-optimized CV models, active learning pipelines, and distributed data workflows to translate ambiguous problems into scalable solutions.

At Google I design, train, and deploy pCVR models for YouTube App Ads, build LLM-powered automation and semantic retrieval systems, and implement low-latency, robust inference for production. I have a proven record of reducing labeling and review costs, improving experiment productivity, and delivering deployable models across cloud and embedded environments while collaborating across global product, infra, and research teams.

Experience

Work history, roles, and key accomplishments

Current

Machine Learning Software Engineer

Current

Google

Jun 2020 - Present (5 years 9 months)

Developed and deployed large-scale pCVR models and NLP systems for YouTube App Ads, improving advertiser ROI and automation workflows while reducing manual experiment setup by ~25–40%. Led end-to-end ML pipelines, bias mitigation, calibration, and A/B validation for ads production.

PCVR CVR Modeling Transformer Models NLP TensorFlow GCP A B Testing Model Monitoring

Senior Data Scientist

Bosch Center for Artificial Intelligence

Aug 2019 - May 2020 (9 months)

Led computer-vision initiatives for driver monitoring, building YOLO-v2 and HourGlass-based eyelid detection models (under 2-pixel RMSE) and deploying optimized models to NVIDIA Xavier edge devices for real-time safety systems.

Computer Vision HourGlass Networks Active Learning Model Quantization NVIDIA Xavier TensorFlow

Data Scientist

Bosch Center for Artificial Intelligence

Jan 2016 - Aug 2019 (3 years 7 months)

Delivered end-to-end ML solutions for manufacturing, including recommendation algorithms for fuel injector settings used by 50+ experts and root-cause analysis engines to reduce part failures.

Recommendation Systems ETL Python R scikit learn Tableau SQL MongoDB

Graduate Research Assistant

University of British Columbia

Apr 2012 - Dec 2015 (3 years 8 months)

Developed high-dimensional mixed-effects and negative binomial models for longitudinal clinical outcomes and implemented performance-critical algorithms in C with R interfaces released as CRAN packages.

Statistical Modeling C R Longitudinal Analysis Biostatistics

Graduate Research / TA Mentor

University of British Columbia

May 2011 - Dec 2015 (4 years 7 months)

Developed robust sparse clustering and non-parametric Bayesian models, mentored graduate instructors, and taught undergraduate statistics courses including STAT 200 and STAT 305.

Clustering Teaching Mentorship Workshops R

Education

Degrees, certifications, and relevant coursework

The University of British Columbia

Doctor of Philosophy, Statistics

2011 - 2016

Activities and societies: Graduate Research Assistant; developed mixed-effects and negative binomial regression models; authored peer-reviewed publications.

Completed doctoral research in statistics, focusing on high-dimensional and longitudinal clinical outcome models and releasing performance-critical implementations and CRAN packages.

The University of British Columbia

Master of Science, Statistics

2009 - 2011

Activities and societies: Graduate Research Assistant / Teaching Assistant; developed robust clustering algorithms and non-parametric Bayesian models; taught undergraduate statistics courses.

Master of Science in Statistics involving advanced statistical modeling and research supporting subsequent doctoral work.