Data Analytics & Operations Leader
Denver, Colorado · open to Denver-area and remote
I spent a decade running restaurants — full P&L, hiring, the floor at peak on a Saturday night — then spent two years teaching myself to build the analytical systems I always wanted to run one with.
Since 2024 I have designed, shipped, and operated two production data systems end to end, alone: data engineering, ML modeling, full-stack development, deployment. My mechanical engineering degree is where the habit started — take an ill-defined problem, find the constraint that actually binds, build the system around it, own what happens next.
Most people have the floor or the model. The combination is the point: I know what an operational decision costs, and I can build the thing that informs it.
A markets-intelligence dashboard I designed, built, and operate end to end. Six modules run in production — market digest, analyst scorecard, sector pulse, baskets, options and global. Intraday capture runs on external cron against authenticated task endpoints rather than in-process schedulers, and a self-grading health endpoint scores every module's freshness with a watchdog that catches staleness before a reader does. The options module carries a live CBOE feed and a scenario tool; an AI pipeline transcribes and grades financial-news segments.
Python · Flask · SQLite · OpenAI API · Render · GitHub Actions · vanilla JS
Calibrated simulation for MLB markets, on a 21-table SQLite warehouse built from 7.8M Statcast pitches across 11 seasons. A LightGBM per-plate-appearance classifier with isotonic calibration feeds a vectorized game simulator — 175× faster than the naive version — producing per-game distributions, with totals and starter workload read out of the same run. I ran a three-month pre-registered forward test, then killed the bet: re-deriving every edge with correctly weighted model and market terms took positive-EV plays from 4,093 to 127, and the blend did not beat the market out of sample. The starter-workload sub-model backtests at 3,886 starts, eight of nine outs lines passing calibration.
Method, data, figures and an interactive explorer are published at lab.dataphil.com ↗ — the negative result with its full audit trail.
Python · SQLite · Parquet · LightGBM · isotonic calibration · pybaseball / Statcast
Founded and built dataphil.com — designing, shipping, and operating production data systems end to end: data engineering, ML modeling, full-stack development, deployment. Both systems above exist because a manual process was costing me time, and the fix was automation rather than more hours. Shipped an MLB game-center dashboard to a podcast partner as a production tool — the first product with a real user who wasn't me. Alongside the build: full-time discretionary US equities trading on my own capital, investing since 2019, with written risk limits and position sizing. The discipline half of the job, and it shows in how I ship software.
A ground-up business build. I joined the opening team, returned as Operations Manager in 2022, and was promoted to Co-General Manager as the restaurant reached profitability within three years of opening — driven by the operational changes my counterpart and I implemented. Our operating model became the template the ownership group used to acquire and scale other restaurants.
I co-owned every lever of a from-scratch operation: product mix and menu, pricing, happy-hour strategy, expansion calls like weekend breakfast, service flow, hiring and training — full P&L, vendor relationships, and brand voice, the ownership stack a corporate parent would normally split across five departments. The closest thing to entrepreneurship without the equity: I helped form the business, shape it, and leave it running.
Full P&L ownership of a $3M+ annual-revenue fine-dining steakhouse with a 40+ person team, on a turnaround mandate in a market I had to learn from scratch — the steakhouse inside Silver Legacy Resort Casino, the largest casino property in Reno, for what was then the largest publicly-traded fine-dining steakhouse chain in the US.
I took the location's aggregate online review scores from bottom-five in the company to above company average in one year — a guest-experience turnaround, not a marketing one — and cut labor cost while holding fine-dining service standards.
I shut the restaurant down when COVID closed the property, then reopened it under protocols stricter than any standalone restaurant faced: inside a casino, gaming-property health and compliance standards stack on top of state and county orders, and I rebuilt service flow, staffing, and training around them. We were one of the only restaurants cleared to reopen with the casino floor, which meant absorbing the property's dining demand and running at full capacity inside those guidelines.
Owned the beverage program end to end — wine list curation, cocktail menu, inventory and cost controls, supplier relationships, bar-staff training and certification. Curating a list is a taste question and a margin question at the same time. When the long-tenured GM left to help open the Denver Tech location, I stepped into running the restaurant in every functional way — schedule, floor, P&L reviews — until the formal promotion to GM in Reno followed.