I use SQL, R, Tidyverse, Excel, and Shiny to investigate operational problems, compare decision options, and communicate evidence. My projects cover seller performance, service-request prioritization, and financial scenarios, with AI-assisted development and explicit validation.
My work begins with the decision behind a metric request, locks the measurement design before execution, validates results with explicit quality checks and documented evidence boundaries, and ends with an evidence-bounded recommendation.
flowchart LR
A["Start and frame"] --> B["Design measurement"]
B --> C["Execute and validate"]
C --> D["Interpret evidence"]
D --> E["Recommend action"]
| Project | What it demonstrates | Core tools |
|---|---|---|
| FulfillIQ 2.0 | Seller-performance simulation: independent SQL and R paths, documented reconciliation across 3,095 sellers, seven investigation candidates, and two unresolved cases | MySQL, SQL, R, Tidyverse |
| Chicago 311 Dispatch Priority | Completed simulation: 222,002 request classifications reconciled across independent R paths; a timezone defect affecting 104 labels caught and repaired; operational capacity limits explicitly recognized | MySQL, SQL, R, validation |
| Five-Stage Analyst Workflow | Framework for moving from a vague stakeholder request to validated evidence and a proportionate recommendation | Decision framing, KPI design, AI quality control |
| AI-Augmented Bitcoin Proxy Analysis | Dilution-aware comparison of six public Bitcoin proxies across three bullish scenarios, with model checks and conditional interpretation | Jupyter, Excel, scenario modeling |
| AI-Augmented Analytics Portfolio | Compact demonstrations of Shiny dashboarding, data-grounded AI reporting, and interpretable classification | R, Shiny, OpenAI API, tidymodels |
Start with FulfillIQ 2.0: case study · SQL/R reconciliation record · final interpretation.
The original FulfillIQ remains available as the Version 1 historical baseline.
- Translate metric requests into explicit business decisions.
- Define hypotheses, KPIs, grain, comparison groups, confounders, and failure conditions before writing SQL.
- Control joins, denominators, nulls, dates, and boundary conditions explicitly.
- Validate important metrics with explicit QA checks, sensitivity analysis, and documented evidence boundaries.
- Use AI models as specialized builders and critics rather than as an unreviewed source of truth.
- Separate facts, interpretation, uncertainty, and recommendations.
- Publish evidence in formats stakeholders can use: Excel, dashboards, reports, and presentations.
Analytics: SQL · MySQL · R · Tidyverse · ggplot2 · tidymodels
Delivery: Excel automation · Shiny · DT · Jupyter · executive reports and presentations
AI integration: OpenAI API · independent multi-model review · analytical quality control
In development: PricePoint, an M5 retail pricing and revenue case study. Database setup is complete; analytical stages have not started. I am also aligning prospective workflow enforcement with the latest measurement and evidence requirements.