Education & experience.

SEP 2022 — JUN 2026  ·  SEATTLE, WA

B.S. Informatics (Data Science)

University of Washington ›

Minor in Economics  ·  GPA 3.62  ·  Dean's List

A data-science-focused informatics degree that covers most of the discipline, from statistical methods and data programming to relational modeling, software practice, and information visualization. The economics minor keeps the technical work tied to how systems and incentives actually behave.

Relevant coursework

JUN 2025 — MAR 2026  ·  SEATTLE, WA

Azure Data Engineering Intern

University of Washington Information Technology (UW-IT) ›

I owned production ETL on Azure from end to end. I designed, built, and maintained scalable pipelines with Azure Data Factory, Databricks, and Synapse Analytics, pulling data together across four operational systems and handing clean, analytics-ready datasets to the teams downstream. The other half of the job was keeping it all running: monitoring production performance, triaging failures, working root causes, and shipping fixes that held SLA-level reliability. I also pushed the platform into AI, designing and deploying agentic RAG solutions on Azure OpenAI Service and the Claude API, from scoping the use case through CI/CD release on Azure DevOps.

Tools used

Lessons learned

• Stakeholders cared more about reliability than about any one new pipeline. Catching 90%+ of issues before they reached anyone downstream is what built the trust.
• An agentic RAG system isn't really production-grade until it has evaluation and CI/CD around it. The model itself is the easy part.
• A dashboard that takes five hours of manual reporting off someone's week earns trust faster than one that just looks impressive.

SEP 2023 — PRESENT  ·  SEATTLE, WA

STEM Remediation & Operations Lead

University of Washington Disability Resources for Students ›

My job is keeping a high-volume estate of production pipelines healthy. I maintain and improve Python ETL across 100+ concurrent workflows each quarter, shipping enhancements and defect fixes that hold a near-perfect on-time delivery rate. I built automated data-quality monitoring that runs completeness, consistency, and anomaly-detection checks at every stage, and I brought the Claude and OpenAI APIs into production as agentic workflows, using tool calling for classification, quality checking, and content routing. I also reverse-engineered and documented a backlog of components nobody had written down, so the team stopped relearning the system from scratch every time.

Tools used

Lessons learned

• Validation belongs at every stage, not just the end. Catching errors early is what drove the 25%+ drop downstream.
• Writing down 10+ undocumented components killed a recurring class of data-handling errors and cut ramp time for new staff. Documentation pays for itself fast.
• AI agents earn their keep on the boring work, the classification, routing, and quality checks, where they gave back about 30% of manual review time.