Principal data scientist · independent

Custom decision systems for operations that outgrew the spreadsheet.

I take messy recurring planning problems — production schedules, inventory allocation, supply optimization — and turn them into clear decision systems your team can run. Not a dashboard. Not a generic optimizer that ignores your constraints.

Book a 30-minute fit call View Selected Work

The problem isn't just “we need more AI.”

Off-the-shelf planning software fails when your problem doesn't fit in their box. Spreadsheets fail when the same complex decision has to be re-solved every day by someone who already has a job.

The problem is usually three things:

1. Data Problems

The numbers are wrong, incomplete, siloed, or arrive after the decision has already been made by hand.

2. Logic Problems

You have the data, but it's a major technical challenge to define what you want and how to get there. Ops and Quality are asking for opposite guardrails.

3. People Problems

The math is right but your teams still struggle to follow through on the plan. Teams held to the wrong KPIs have misaligned incentives. Accountability and decision authority is unclear.

My Solution

A decision system you and I define together that knows your data, your goals & guardrails, and your organization.

How we will work together

We will begin with a brief conversation to ensure the problem is a fit. Then assess which plan(s) work best.

1–2 weeks

Diagnostic

Interview key team members, review tech stack and data architecture, clarify goals and pain points. You get an assessment on readiness—data, logic, org—plus a sharpened success metric and a recommendation: stop, run a prototype, or scope a fixed build.

Time-boxed

Prototype

A working but limited system built on your real constraints — enough to see concretely whether you require a production system, before you fund the full build.

Scoped build

Production System

A one-time build of the system you and I define together: scheduling, allocation, blending, supply planning and more. Sized to fit your tech stack and team.

Monthly

Retainer or Fractional Lead

Keep a system healthy and nimble after launch. Provide a junior data team with principal-level architecture and oversight without a full-time hire.

Selected Work

From 10+ year employment history in data science & analytics.

80 Acres Farms · Operations

Production Schedule Stabilization

The Problem

  • Daily crop scheduling had complex time and space constraints
  • Propagation area was the bottleneck and frequently overflowed
  • Day-to-day schedules highly volatile; broke labor, harvest, and fulfillment

My Solution

  • Mapped the existing scheduling method, simulated the farm material flow
  • Shipped a constrained optimizer that cut volatility while respecting capacity and crop rules

The Result

  • Daily schedule volatility down 80%
  • Space bottleneck overflows became rare exception
  • Downstream S&OP got a signal it could trust

80 Acres Farms · S&OP

Product Blending and Order Allocation

The Problem

  • Complex daily production planning due to short shelf life, volatile demand, flexible BOM
  • Planners waited for after-hours data
  • Two error-prone hours of data input, trial and error tinkering, unofficial rules

My Solution

  • Codified constraints & goals with operations, sales, and quality
  • Deployed custom decision system that recomputed daily optimal blends
  • Allocated scarce goods inventory with explicit customer prioritization

The Result

  • Head of S&OP daily planning time reduced from 2+ hours to 20 minutes
  • Better fulfillment and less waste
  • Siloed tribal knowledge became codified and integrated

Kroger · digital

New-to-You Product Recommendations

The Problem

  • Kroger website could only recommend repeat purchases
  • Good for replenishment. Weak for basket growth
  • Standard algorithms based on one-to-one affinity gave mediocre results

My Solution

  • Novel approach took word embedding techniques from language-modeling and adapted them to finished goods
  • Detected richer relationships among finished goods than possible with standard algorithms

The Result

  • Homepage offered personalized new-to-you product recommendations for the first time ever

Philosophy & Background

I have spent more than ten years building production data science solutions in vertical farming, grocery, and fintech. In that time I've seen that simplistic solutions provide math and models in isolation, whereas simple solutions provide them in the proper context of data, people, and processes. My creative energy comes not from just training another machine learning model, but from taking a complex knotted operating decision and untying it into a clear simple solution.

I will not sell you hype, fear, uncertainty, or doubt. There are times when the shiny AI tool is the right solution, and plenty of times when it is not. Staying grounded in reality gives the best ROI.

Background that shapes the work: I am “whole brained.” A right brained person by nature –philosophy major, avid reader, hobbit enthusiast—who about 15 years ago fell in love with the creativity and capabilities of technology and math.

Tell me the decision you keep re-solving.

A 30–60 minute call to see if it's a good fit. If it is, the next step is usually a 1–2 week diagnostic, not a six-month proposal.

A fit call covers topics such as: