James Poindexter · Orlando, Florida

I ran the operation.
Then I built the system that runs it.

Fifteen years managing real operations — commercial kitchens, P&L, vendors, federal contracts, community management — then I taught myself to build software with AI leverage. Now I build systems small businesses own outright: no per-seat subscriptions, no consultant hostage-taking, numbers you can trust.

$9.85cost per plate, computed to the penny through a 5-component recipe tree
136dishes cataloged, filterable by cuisine × meal × course
13tables in the vendor procurement & scoring module
0numbers shown that the system can't prove — the honesty rule

How I build

Three rules, enforced in code

These aren't slogans — every system I ship is built around them, and I can show you the code that enforces each one.

Store what happened, derive the judgment.

The system records what a vendor promised and what actually arrived — never a typed-in opinion. Change the standards later and all history re-scores itself. Judgments never go stale.

AI drafts, humans approve.

AI does the tedious part — reading emails, extracting receipts, drafting records — and everything it produces waits in a review queue. Nothing becomes business truth until a person accepts it.

Honest numbers or no numbers.

If one ingredient in a recipe is unpriced, the dish shows no cost — not a guess. A blank you investigate beats a wrong number you quote to a customer.

Skills map

What I bring, by area

AreaWhat I doProven in
Operations management
P&L ownershipVendor managementProcurementFederal contract mgmtCommercial kitchensCommunity managementProcess design
15+ years running operations before ever writing code — I speak owner, not developer.
Every system below runs a real business — mine.
Systems & data
Data modelingSQL / SQLiteCosting enginesPipeline designReportingMigration with audit trails
One database as the single source of truth; every number drillable to the rows behind it.
Catering operations system; vendor scorecard
AI integration
Voice-first data captureDocument & email extractionDraft-and-review workflowsKnowing where AI doesn't belong
AI where messy language lives; deterministic code where money lives. Never the reverse.
Voice capture pipeline; lead-to-proposal flow
Tools
PythonSQLJavaScriptCloudflare (D1, Workers)Git / GitHubClaude & LLM APIs
Deliberately boring, durable stack — cheap to run, easy to hand over, yours to keep.
All of the below

Proof of work

Systems in production

Built for my own catering company, Epic Curiosity LLC — which means I was the client, the user, and the builder, and the system had to survive real service.

Catering operations system

Private repo · in production

The full back office for a catering business in one system: a recipe costing engine that computes plate cost through nested sub-recipes; a 136-dish catalogue tagged by cuisine, meal, and course; an 8-stage booking pipeline where payment status is computed from money actually recorded — never typed; and a finance screen where every total opens into the rows it came from.

Sample output: Elevated Chicken Paprikas · 30 lines · 5 sub-recipes · $39.39/batch · $9.85/plate · 0 blocked lines — and if one line couldn't be priced, it would refuse to show a total at all.

Voice-to-database capture

AI + human review

Recipes, supplier notes, and business decisions captured by voice — while driving — straight into a cloud database. A review bench then makes the vague precise: every AI-drafted line is flagged, and nothing reaches the books until a person accepts it. The system refuses to promote a guess.

The pattern that matters for your business: capture anywhere → AI drafts → you approve → system of record. Works for receipts, leads, and orders just as well as recipes.

Vendor scorecard

Public — read the code

A procurement pipeline that scores vendors on what actually happened: quotes vs. invoices, promised windows vs. actual deliveries, silence as a scoreable event. Thresholds are configuration, so tightening your standards re-scores all history. Open source on GitHub with a runnable demo.

The demo's punchline: the vendor with the cheapest quote scored 56/100 overall with delivery at 26.7 — invoice cost is not true cost, made visible in one row.

Lead-to-proposal flow

AI + human review

Paste a marketplace lead email and 17 fields pre-fill; a proposal builder shows the client's ask next to your catalogue, checks coverage and budget honestly ("$3.20 OVER their band" — never a silent overrun), and produces a client-ready proposal sheet. Code guarantees no internal cost ever appears on a client-facing page.

Working together

A fixed-price ladder — each step de-risks the next

No hourly billing, no open-ended retainers. You know the price before the work starts, and you walk away with something real at every stage — even if you stop there.

StageYour costYou walk away with
Discovery call Free · 30 min A straight answer on whether a system would pay for itself — and one useful observation either way.
Systems audit $750–1,500 fixed100% credited if you build A map of how your operation actually runs, the bottlenecks ranked by payoff, and a fixed-price build proposal — yours to keep even if you never call me again.
Implementation Fixed price from a published price booktypical range $2,500–12,000 A working system, your data migrated in (every ambiguity reported, nothing invented), a trained team, written runbooks — and you own all of it.
Care plan Optional monthly feetwo simple tiers Backups, updates, fixes, and steady improvements. Convenience, not captivity — you own the system either way.
Loyalty Built in Existing clients get 10–15% off new projects; referrals earn credit.

Does this sound like you?

Five signs your software doesn't fit

Most owners I talk to have at least two of these. If you have three, a system would pay for itself inside a year — and I'll tell you honestly if it wouldn't.

Every one of those is a bridge you built because the software wouldn't bend. You're paying twice — once for the subscription, again for the hours spent working around it.

Contact

Drowning in spreadsheets? Let's talk.

First conversation is free, 30 minutes, and worth your time even if we never work together. Catering and food businesses are my home turf; any small operation buried in double-entry is a fit.