Data scientist focused on operational decisions, optimization, and evaluated AI.

Data and AI for decisions that need to hold up.

I use statistical analysis, optimization, and evaluated AI to help teams make operational and product decisions with evidence they can defend.

Open to senior data science, applied AI, and analytics roles
Toledo, Ohio Remote, hybrid, or on site

Python
SQL
Statistical inference
Constraint optimization
AI evaluation
Decision dashboards

Twenty measured results, including seven that went against the build. Both favorable and unfavorable findings are published with the method. Review the evidence ledger

View resume View selected work

10+ years: Operational ownership
Approx. 25: Employees led
8 apps: Live portfolio work

Black Box AI findings page summarizing 7,462 public aviation safety reports and fatality patterns — Evaluated retrieval NTSB public data
Black Box AI

**My role:** Independent evaluated AI project. I designed the question routing, guarded SQL path, retrieval evaluation, answer evidence, and audit trail.

**Result:** Hybrid retrieval did not clearly separate from simpler methods, so the product defaults to semantic search and exposes all three for comparison.

Hybrid MRR
0.902

Semantic MRR
0.853

Decision
No clear win

Pilot evaluation set: 17 hand-labeled questions. The set is too small to support a method choice.
Read the case study

Selected work

Optimization and statistical analysis.

Each case states the decision, method, ownership, result, and limit. Full implementation detail remains on the case-study page.

View all projects

Decision optimization Synthetic benchmark

Dispatch Optimizer

Recommend feasible technician assignments under skills, capacity, service deadlines, travel, and overtime constraints.

Method: OR-Tools CP-SAT vehicle routing with time windows plus assignment, compared with naive and greedy plus 2-opt across 200 randomized scenarios.
My role: I defined the synthetic operation, implemented both baselines and the optimizer, and built the comparison interface.
Result: Greedy plus 2-opt beat naive in 199 of 200 scenarios, with a median 12 fewer service-level breaches. Across 24 scenarios, CP-SAT's median was 5 breaches behind that baseline at 8 seconds, 3 ahead at 30, and 5 ahead at 60.
Limit: Synthetic data, not a production operation. The timed solver is nondeterministic, so the distribution matters more than any one scenario.

Read case study Open live demo

Synthetic dispatch benchmark comparing a baseline plan against an optimized plan on the same crew, jobs, and constraints — Synthetic benchmarkSame crew, jobs, and constraints

Statistical inference Public mortgage data

Fair Lending Lab

Help analysts identify disparities that merit deeper review without turning association into a causal claim.

Method: Five preregistered hypotheses on 41,287 curated applications from 210,643 raw HMDA rows, using effect sizes, confidence intervals, resampling, Bayesian sensitivity, and multiplicity controls.
My role: I defined the hypotheses, implemented the statistical and sensitivity tests, and built the analyst dashboard.
Result: The raw denial-rate gap was 10.7 percentage points. It fell to 4.73 after financial controls, with 44.4 percent surviving adjustment.
Limit: HMDA lacks key underwriting variables. These findings are screening signals for deeper review, not discrimination findings.

Read case study Open live demo

Fair Lending Lab dashboard showing a lead finding, effect size with confidence interval, and five corrected hypothesis tests — Public-data analysis41,287 curated applications

Operating record

I came to data through operations.

My technical work is grounded in accountability for people, workflows, data quality, and decisions. That context shapes what I measure and what I refuse to overclaim.

Practice leadership: Founded and ran a behavioral health practice that grew to approximately 25 employees, with responsibility for staffing, payroll, billing, vendors, compliance, and financial reporting.
Manufacturing analytics: Build SQL data models, reporting workflows, and management-ready dashboards close to a high-volume production line.
Technical delivery: Eight live portfolio applications across statistical inference, optimization, retrieval, recommendation, and decision dashboards.
Education: M.S. Data Science, Eastern University, in progress. M.S. Information Systems Management, University of Arizona Global Campus, GPA 4.0.

Capabilities

How I contribute.

Applied AI evaluation and analytics

Frame viable use cases, define success measures, evaluate retrieval and model behavior, and design reviewable safeguards.

Statistical analysis and optimization

Use inference, forecasting, scoring, or constraint models to answer a named operational or product question.

Decision dashboards

Turn validated findings into concise tools that leaders and operators can inspect and use.

Read how I frame the work

Senior data science and applied AI roles

Senior data science, applied AI, and analytics work.

I am open to senior roles where analytical rigor, operational context, and trustworthy delivery matter.

View resume Contact Scott