Open to opportunities

Yongwoo Noh

@yongwoonoh

Senior machine learning engineer at Pinterest who built an end-to-end board recommendation pipeline.

At Pinterest, I built an end-to-end board recommendation pipeline spanning graph-embedding retrieval, a PyTorch ranker, and production serving. I also integrated long-term user history with real-time session signals and added fallback retrieval for sparse-history and seasonal users.

At Block / Cash App, I owned search ranking and the ML inference platform. I trained XGBoost and Transformer rerankers evaluated on payment completions, and unified model deployment through Docker and Kubernetes.

Earlier, I improved Home Timeline ranking at Twitter, built job-search ranking and indexing systems at Google, and worked on LinkedIn Job Search recommendations and Galene search infrastructure. Across these roles, I’ve worked on ranking models, retrieval, training pipelines, and ML serving systems.

Experience

Work history, roles, and key accomplishments

Current

Senior Machine Learning Engineer

Current

Jan 2024 - Present (2 years 9 months)

Built end-to-end board recommendation pipeline including graph embedding retrieval, ANN candidate generation, multi-task PyTorch ranker, and serving. Integrated lifelong user history with real-time session streams and ran A/B experiments to improve recommendation quality.

PyTorch Faiss PinSage GNN Spark Kafka Apache Flink Statsig Transformers GNN Feature Stores A B Testing Recommendation Systems

Senior Machine Learning Engineer

Block / Cash App

Jan 2022 - Jan 2024 (2 years)

Owned Cash App search ranking and ML inference platform serving the entire ML org. Trained XGBoost and Transformer re-rankers for payee search and architected unified deployment for multiple model families.

XGBoost Transformers Docker Kubernetes AWS Sagemaker Metaflow MLFlow gRPC Snowflake DBT Datadog Statsig Feature Stores

Senior Machine Learning Engineer

Twitter

Jan 2020 - Jan 2022 (2 years)

Improved Home Timeline heavy ranker features using SimClusters, GraphJet, and engagement history. Rebuilt training pipelines and automated model refresh, cutting model staleness from ~7 days to ~1 day.

SimClusters GraphJet Kafka Hive Cortex Starlark A B Testing Ranking Feature Engineering

Software Engineer

Google

Jan 2017 - Jan 2020 (3 years)

Built query processing, job indexing, ML ranking integration, and gRPC response serving for Cloud Talent Solution. Owned title, seniority, and industry normalization pipeline and shipped API to production across 3,000+ job properties.

NLP Learning To Rank BigQuery gRPC Protocol Buffers Search Classification

Senior Software Engineer - Machine Learning / Data Mining

Jan 2013 - Jan 2017 (4 years)

Operated across ML ranking for Job Search and Job Recommendations, and backend platform for Galene search-as-a-service. Trained GBDT and GLMix personalization models on professional graph features and extended Galene with vertical-specific ranking APIs.