Data Analyst — SQL · Python · Experimentation · Decision Analytics
MS in Business Analytics, Carlson School of Management, University of Minnesota Minneapolis, Minnesota · Open to Data Analyst roles in any industry
I turn messy data into decisions. Most analyses stop at a chart; I try to finish the sentence — what changed, why, what it's worth, and what we should do about it.
What I enjoy most is the part before the model: finding out that a spike was duplicate records, that a feature was only knowable after the fact, or that a metric everyone quotes was never defined the same way twice. Getting that right is usually worth more than a better algorithm, and it's the difference between an analysis people act on and one they quietly ignore.
How I work: validate before explaining · write the definition down · answer in decisions, not dashboards · state what would break the conclusion.
| Project | Question it answers | Result |
|---|---|---|
| MetricGuard AI | Why did the dispute rate spike? | A third of a 32.8% spike was duplicate data; the real driver was travel on mobile |
| Retail Customer Analytics · live dashboard | Where should the retention budget go? | 335 slipping high-value customers, £2.0M of 12-month revenue at stake |
| Fair & Explainable Credit Risk | Who should get credit, and is the model defensible? | Cost-based threshold cut the approved-book default rate from 15.7% to 9.4% |
| Bank Marketing Targeting | Which customers should the call center call? | 87% of conversions from 60% of calls; 4,117 calls saved |
| Experimentation & Causal Impact | When is the obvious read wrong? | Paid-search ROI fell from 320% to 80% once organic substitution was removed |
| Restaurant Recommender | Which recommender should ship? | Matrix factorization for warm users, popularity fallback for cold start |
Each repo runs from a clean clone, has automated tests and continuous integration, and states its own limitations.
Oct 2025 – Aug 2026
C.H. Robinson (Fortune 500 logistics)
- Built a KNN matching model estimating relationship health for the 54% of accounts that never respond to the customer survey, extending churn early-warning coverage to all 22K active accounts.
- Designed a two-axis prioritization matrix combining revenue-weighted growth and comment sentiment, flagging ~$6B of current revenue as satisfied-but-shrinking for retention outreach.
- Delivered a SQL-to-Python pipeline and dashboard giving VPs and marketing and operations stakeholders a standing retention-priority view; served as Scrum Master and Product Owner for a 5-person team.
Oct 2025 – Aug 2026
4Mativ (school-transportation technology)
- Engineered a GPS data-quality layer over 2.35M pings and 10,800 trips, scoring connection health with per-route Isolation Forest models and isolating a device fault affecting 19% of one provider's trips against under 2% elsewhere.
- Surfaced 1,191 detour and 248 wrong-route trips with DBSCAN corridor modeling, concentrating 64% of service failures in 3 of 16 vendors.
- Built a vendor reliability matrix pairing GPS health against operational execution, separating fleets with broken tracking from fleets with real service failures so each got the right fix.
Oct 2025 – Aug 2026
Central Specialties (Midwest infrastructure)
- Constructed the bid-level dataset behind a competitive-intelligence engagement — 2,736 bids, 629 projects, 419 rival firms — and defined 5 competitor KPIs across market presence, win share, success rate, pricing aggression, and winning margin.
- Sized $15.6M of margin left on won projects and modeled a 3% price reduction that would raise win rate 43% across $223M of near-miss contract value.
- Presented an interactive Tableau dashboard with county-level competitor mapping, letting the bidding team price against each rival's historical behavior before committing estimating resources.
Jun–Sep 2024
- Extracted company financials through the Wind API and consolidated three vendor sources into 12 analysis-ready tables covering 100+ manufacturers, building the comparative valuation base for a new coverage segment.
- Built top-down market-sizing models under multiple growth scenarios, producing the revenue forecasts behind the firm's first two published reports on the segment.
Jun–Sep 2023
- Screened 200+ employee financial-disclosure and account records against internal compliance rules, reconciling data across systems to identify undisclosed holdings and conflicts of interest.
- Coordinated across retail banking, compliance, and corporate relationship teams to onboard payroll accounts for 5 enterprise clients and 300+ employees, performing KYC verification for fraud and account-misuse risk.
Languages: SQL · Python · R
Analysis: A/B testing · causal inference (DiD, propensity matching) · segmentation & RFM · cohort and retention analysis · classification · lift and ROI analysis · anomaly detection · model explainability and fairness auditing
Stack: pandas · DuckDB · scikit-learn · XGBoost · statsmodels · Tableau · Streamlit · Plotly · Git · GitHub Actions · pytest