I build decision-support systems with SQL, R, Tidyverse, Excel, Shiny, interpretable modeling, and AI-assisted analytical 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 | Seller-performance decision case: stakeholder framing, locked KPI design, a reviewed MySQL analysis specification, R validation of the committed seller export, Excel evidence, three-AI review, and an operational recommendation | MySQL, SQL, R, Tidyverse, Excel |
| AI-Augmented Bitcoin Proxy Analysis | Dilution-aware comparison of six public Bitcoin proxies across three scenarios, with reproducible model checks and decision-focused interpretation | Jupyter, Excel, scenario modeling, validation |
| Five-Stage Analyst Workflow | A complete framework for moving from a vague stakeholder request to validated evidence and a proportionate recommendation | Decision framing, KPI design, AI quality control |
| R Workflow Engine | A one-file specification for generating structured tidyverse workflows with preparation, analysis, assurance, and publication stages | R, Tidyverse, Excel, Shiny |
| AI-Augmented Analytics Portfolio | Compact demonstrations of Shiny dashboarding, data-grounded AI reporting, and interpretable classification | R, Shiny, OpenAI API, tidymodels |
- 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
I am developing practical analytics systems that combine rigorous measurement, reproducible implementation, AI-assisted review, and clear decision support.