Work

Leading master data management for Thrasio's catalog operations: I own the NetSuite item and channel data model, automate SKU workflows, build the team's BI reporting stack, and set data quality standards across NetSuite and downstream systems. I work cross-functionally with catalog, sourcing, supply chain, and tech, and mentor teammates on automation and analytics tooling.

• Rebuilt the SKU creation and update workflow as a full-stack Streamlit app built with Gemini CLI as an AI coding agent, cutting the average SLA from 24 hours to 30 minutes (a 95% reduction) and reducing a two-person job to one person.
• Automated multi-step data validation and NetSuite integration, eliminating an estimated 5 hours of manual spreadsheet work per analyst per week.
• Implemented Google OAuth authentication for internal tooling, enabling secure, role-based access for MDM team members.
• Led the rebuild of NetSuite's item and channel data model, using Gemini CLI to reverse-engineer undocumented legacy tooling before restoring a single source of truth for catalog data.
• Authored and rewrote 3 MDM SOPs (SKU creation/update, dimension & weight validation, 3PL master lookup), taking dimension/weight validation to 100% coverage across all 2025 outbound POs.
• Built the team's Power BI and Looker Studio KPI dashboards (SKU maintenance, Amazon Order-to-Cash, Stord monthly transactions) and led the team's migration from Power BI to Looker Studio.
• Ran internal training sessions on Google Apps Script automation, Looker Studio, and GitHub that helped teammates build their own data pipelines and dashboards, cutting out a recurring manual download/upload task.

Supply Chain Business Analyst responsible for BI dashboards, ad hoc reporting, and data quality initiatives supporting the purchasing, planning, and warehouse teams.

• Built Power BI dashboards for the purchasing team covering Inventory Stocking Levels and Purchase Order Tracking, using Azure SQL and Business Central Web Services as data sources.
• Reduced report refresh time by 50% by standardizing data fetching methods and query patterns across the BI layer.
• Maintained a 95% on-time delivery rate for purchase orders by coordinating across marketing, planning, and warehouse teams.
• Identified and corrected KPI calculation errors in presale and supply chain reports, contributing to approximately $70,000 in annual cost savings.

Managed organizational databases, maintained data integrity across client records, and provided technical support to operations and sales teams using MS Office and ERP systems.

• Built operational and sales reports in Excel and ERP, reducing manual lookup time.
• Modeled company-specific trends using SQL, Excel, and Python on internal datasets, improving data processing efficiency by 10%.

Provided bilingual (English/Mandarin) service to over 100 donors and participants by email and in person at the 2023 GVA Relay for Life, Swangard Stadium, Burnaby.

• Coordinated setup and participant support for the 2023 GVA Relay for Life at Swangard Stadium, Central Park, Burnaby.
• Translated event processes and logistics for participants in both English and Mandarin.

Physical Andon-style status light for Claude Code sessions, driven automatically by Claude Code's hook system, so I can walk away from the terminal and still know at a glance whether the agent is working, waiting, or stuck.

• Designed a 5-state wire protocol and firmware watchdog (RP2040/C++) that fails safe to a distinct "stale" color instead of silently freezing if the host driver ever crashes.
• Integrated with 8 Claude Code hook events end-to-end, including a permission-request-specific state and a dedicated "compacting" color so the light never misrepresents what the agent is actually doing.
• Found and fixed a race condition caused by async hook execution, where a delayed background command could silently overwrite a later, correct state — a bug only visible under real session load.
• Took the project from kickoff to a fully working breadboard MVP (firmware, host CLI, and Claude Code hooks all validated end-to-end on real hardware) solo, in under a week.

Statically generated portfolio using Next.js 16 App Router, Tailwind v4, shadcn/ui, MagicUI, and @react-pdf/renderer for server-side PDF generation.

Streamlit web app that takes a SKU update or creation request from validation through NetSuite updates to email notification in a single workflow, replacing a fragmented multi-step spreadsheet process.

• Architected and built the app with Gemini CLI as an AI coding agent, covering design, implementation, and code review, learning firsthand where agentic coding needs human oversight.
• Won 1st place in Thrasio's Automation Olympics, an internal competition for team-built automation tools.

Led a cross-functional project with NetSuite administrators and the tech team to consolidate item data that had fragmented across multiple deprecated tools and team-owned mapping spreadsheets. Authored the ERD-to-NetSuite mapping analysis, implementation plan, and business rules for a suite of custom records — Channel Listing, Item Channel Listing (junction), Fulfillment Node, and a Brand → Category → Group → Type classification hierarchy — plus native record customizations for Item, Customer, and Department.

• Designed the ERD and translated it into a full NetSuite implementation plan, including native record customizations and 7 new custom record types.
• Recovered and re-implemented data structures from deprecated internal tooling, using Gemini CLI to reverse-engineer undocumented legacy code and trace critical but unofficial data sources before restoring lost functionality.
• Documented field-level business rules and data ownership for every custom record, defining how catalog, brand management, and sales teams maintain the data going forward.

Designed and built Power BI dashboards for Supply Chain Operations Excellence covering MDM SKU maintenance KPIs, the Amazon Order-to-Cash transaction compliance process, and Stord monthly 3PL transactions, then led the team's migration of these dashboards to Looker Studio as part of a company-wide tooling change. Also built a Python/Snowflake pipeline that pushes ad hoc data extracts to Google Sheets on demand, replacing manual SQL exports for one-off reporting requests.

• Delivered 3 production Power BI dashboards covering MDM SKU KPIs, Amazon Order-to-Cash compliance, and Stord monthly transactions.
• Migrated the team's BI reporting from Power BI to Looker Studio, standardizing on table blending and cross-filtering to keep queries maintainable.
• Built a Python GUI tool that pulls data from Snowflake (dbt-modeled marts) and pushes it to Google Sheets automatically for ad hoc reporting requests.

Led a set of data-quality initiatives addressing inconsistent SKU naming, incomplete item attribute validation, and inaccurate HTS (Harmonized Tariff Schedule) codes across NetSuite's active SKU catalog. Built Python scripts to review and standardize SKU naming formats, defined updated attribute validation criteria that account for packaging-type variability, and built the review workflow that reconciles NetSuite HTS codes against QA-validated values.

• Standardized SKU naming conventions across the active catalog, replacing an inconsistently followed process with a structured, automatable format.
• Defined updated item attribute validation criteria covering packaging-type variability, adopted as the new MDM validation standard.
• Built a review workflow to reconcile NetSuite HTS codes against QA-validated values, filtering out already-correct, unreviewed, and duplicate SKUs to focus updates only where needed.

A Streamlit app that parses raw Star Citizen Game.log files and generates meaningful statistics and event summaries, helping players quickly review session performance without digging through log files manually.

• Parsed and surfaced play session KPIs including session duration, death counts, and travel logs.
• Enabled players to instantly identify the cause of death for each in-game event.

Collaborated with data analysts from Vancouver General Hospital to develop predictive clustering models forecasting CDI outbreaks from canine screening data collected between 2017 and 2022.

• Built CDI forecasting models in R using clustering and correlation analysis on 5 years of canine screening and infection records.
• Delivered exploratory and risk analysis in Tableau, surfacing correlations between canine alert frequency and CDI case counts.

Used the Box-Jenkins method and dynamic regression approaches on SteamDB data (2020–2022) to fit a multiplicative seasonal ARIMA model and produce a 6-month server utilization forecast.

• Fit a seasonal ARIMA(2,1,2) × (0,1,1) model with a 7-day period using R.
• Forecast 9,487,428 peak concurrent players against an actual of 9,484,341, a 99.97% accuracy rate.

Hackathon project using geospatial clustering to identify optimal locations for new vaccination sites, drawing on 20 census datasets from the Hong Kong government and 5 independent sources.

• Compiled and cleaned 20 relational census datasets from the Hong Kong government and 5 independent sources using R and Alteryx, saving the team 10 hours of data preparation.
• Streamlined geospatial visualizations in Tableau, saving the project team approximately 2 hours per day during development.

Analyzed amenity density and distribution across the Greater Vancouver Area using OpenStreetMap data to build a predictive model for average regional household income.

Coursework in statistical modelling, machine learning, data systems, and applied mathematics. Completed capstone and research projects in collaboration with industry and healthcare partners.

• Dean's List 2021
• Vancouver Coastal Health: Canines-for-Care (CDI forecasting with VGH data analysts)
• Steam Server Usage Forecast (seasonal ARIMA time-series study)