Senior data analyst with 7+ years turning messy multi-system data into trusted KPIs, Tableau dashboards, and decisions for public-sector programs serving 1M+ residents.資深數據分析師，七年以上經驗。把散落在多個系統的雜亂資料，整理成可信的 KPI、Tableau 儀表板與決策依據，服務超過百萬居民的公部門計畫。
Finds the question nobody asked, and pushes back on scope when a request would mislead.找出沒有人問的那個問題；當需求會產生誤導時，會提出異議。
Simulated a proposed 20-to-4 category rollup and found it would understate inpatient per-person cost by about 70%. Pushed back with an 8-category alternative plus a lookup table. Budget error rate fell 12%.模擬把 20 個類別合併成 4 個的提案，發現住院人均成本會被低估約 70%。我提出異議，改提 8 個類別加對照表的方案。預算錯誤率下降 12%。
Traced a six-year collapse in OUD diagnosis records to a 2020 coding-policy change. Negotiated a HIPAA-compliant re-linkage method that restored about 3,000 records.追查 OUD 診斷紀錄六年間的驟減，找到原因是 2020 年的編碼政策變更。協調出符合 HIPAA 的重新串接方法，找回約 3,000 筆紀錄。
Reconciled 6,300 children’s cases across two disconnected systems to show a media-reported autism surge was duplicate counting. The public Tableau dashboard was cited by reporters.比對兩個互不相連系統中的 6,300 件兒童個案，證明媒體報導的自閉症個案暴增其實是重複計算。公開的 Tableau 儀表板獲記者引用。
Built four evidence-based clinical indicators, validated within ±2% on about 200K members over 10 years, in partnership with University of Pittsburgh clinical experts.與匹茲堡大學臨床專家合作，建立四項實證臨床指標，以約 20 萬名會員、10 年的資料驗證，誤差在 ±2% 以內。
Automated the reporting workflow and cut ML feature extraction ETL time 80%, moving the data science team from weekly to daily model iteration.將報表流程自動化，並把 ML 特徵擷取的 ETL 時間縮短 80%，讓資料科學團隊的模型迭代從每週變成每天。
Credited as Ping Jui Ko in the Author/Analytics list on the last page of the report PDF. Built the analysis with my team lead for this Allegheny County DHS report. It covers more than 3,500 children who lost a parent to an overdose from 2016 to 2023, and their use of services before and after the loss.報告 PDF 最後一頁的 Author/Analytics 名單列有我的名字 (Ping Jui Ko)。與 team lead 一起完成這份 Allegheny County DHS 報告的分析。報告涵蓋 2016 到 2023 年間超過 3,500 名因藥物過量失去父母的子女，以及他們在失去父母前後使用服務的情況。
Worked with the University of Pittsburgh Medicaid Research Center and the DHS data warehouse team to add the CMS Chronic Conditions Data Warehouse (CCW) condition definitions to this public county tool. Anyone can now count clients by chronic condition.與匹茲堡大學 Medicaid Research Center 和 DHS 內部的 data warehouse 團隊合作，把 CMS Chronic Conditions Data Warehouse (CCW) 的慢性病定義寫進這個郡政府公開工具。任何人都可以依慢性病別查詢個案數。
A job monitor I built with AI coding agents. It checks company career pages on a schedule, stores the results in Supabase, scores each role against my profile, and sends the matches to Telegram. It runs on GitHub Actions, so there is no server to maintain.我用 AI coding agent 做的職缺雷達。它定時查看公司官網的職缺，把結果存進 Supabase，依我的條件替每個職缺評分，再用 Telegram 通知我。排程跑在 GitHub Actions，不需要自己養一台伺服器。