Control plane
How the work actually ships
Hover a node — click to jump
Montevideo, Uruguay · Cirrus Aircraft · Private Aviation
Kevin Hsueh
Business Data Analyst
MDM · Lakehouse · Applied AI · Analytics Engineering
I modernize enterprise data
so teams stay in the build zone —
not fighting manual ops. and automate operational overhead so teams stay in the build zone — not fighting manual ops.
Profisee MDM — Admin through Developer
All on CV →Practice
Four domains. One operating pattern.
Analysis sets the question. Architecture holds the truth. Engineering ships the system. Applied AI sits on top of metadata — not instead of it.
Business analysis
Turn messy operational questions into decisions leadership can act on.
- Retention models across 40+ BPO accounts
- Org design — spans, layers, headcount scenarios
- WFM capacity planning and forensic billing recovery
Architecture
A single source of truth, with governance that survives the next system.
- Modular MDM on Profisee (LP1–LP4)
- Microsoft Purview data governance
- Databricks lakehouse, medallion lineage, star schemas
Engineering
Pipelines and frontends that stay cheap to run — no database to babysit.
- Python ETL into Parquet star schemas on R2
- Cloudflare Pages Functions at the edge
- Native React dashboards over signed data
Applied AI
Assistants grounded in metadata and lineage — not a chatbot bolted on.
- Glossary recommendations from Purview / Databricks
- Structured Jira stories from freeform notes
- Usage-cost monitoring and provider-to-provider migration
Cirrus work isn't demoed. Public builds below use the same pattern.
Proof
Selected work
Public systems that mirror the day-job pattern: observe the pipeline, model the decision, ship the interface.
Pipeline observability
GitHub Actions Dashboard
Problem: CI health lived in a dozen repos. Checking it meant hopping.
Outcome: One view — run trends, success rates, and workflows that need attention. Edge-cached GitHub REST via Pages Functions.
GitHub APICloudflare FunctionsReactRecharts
Analytics delivery demo
Listener Analytics
End-to-end when I own the stack — Python ETL, Parquet star schema, React frontend over 6+ years of listening history.
PythonParquet / R2ReactRecharts
Initializes the live environment by downloading compressed analytics data.
Enterprise BA
US Attrition
Retention across 40+ BPO accounts. ADF into a star schema, DAX for cohort and forecast — anonymized for public viewing.
Power BIAzure Data FactoryDAX
Decision models
Gamer Analytics
One star schema across Steam, GW2, MapleStory, and BDO. Marketplace preorder ranking under scarcity — the same constrained-decision muscle as capacity planning.
Python ETLParquet / R2React
Full record on the CV. Decisions in the case studies.
Role history, credentials, and the write-ups behind the dashboards.