Work · Consulted by Kenough

Engagements

client work · analytics · full-stack web

Green Sparrow Tattoo Co. / Toasted Bean

Booking, intake, and owner-readable reporting

Direct work with two local owners: booking and intake flows, client communication support, dashboards, and reporting views they can actually read. Scoped in conversation, iterated around usability. The studio runs it on their account, their domain, their data.

Weekly booking requests went from about one to about four inside the first month, and settled around five new client bookings a month across ten months, with roughly 83 percent of intakes arriving complete enough to act on without a follow-up.

source: green sparrow booking records, may 2025 to february 2026, about 96 requests total

tools: Next.js · Supabase · Streamlit · PostgreSQL · SQL

read the full write-up ↗

applied ai · operations

Unified Transactions

Contract-to-close workflow for real estate, deployed

Transaction coordinators juggle deadlines, documents, and client questions across fragmented tools. I modeled the contract-to-close workflow and built it: a Next.js front end over a FastAPI service that extracts the details from contract documents and drafts the updates, with humans approving everything that matters. Deployed on infrastructure I stood up myself.

tools: Next.js · React · FastAPI · SQLModel · LangChain · Claude API · Postgres · Terraform · AWS

read the full write-up ↗

Built outside client work

Not client engagements, and separated from them for that reason. This came out of ASU's MS in AI in Business, built with a team rather than for a paying owner.

graduate program · applied ai · analytics

MainStreet Copilot

Audit checks with the evidence attached

Small businesses have useful data trapped in CSV exports and invoices. MainStreet ingests them, runs deterministic audit checks for duplicate invoices and price variance, and explains findings in plain language with the evidence attached. Built with a team, and the reason the ownership rule on this site exists: the hard part was never the model, it was making the output reviewable by the person whose work depended on it.

tools: Python · SQL · FastAPI · SQLite · OpenAI

read the full write-up ↗

The record

local businesses running software I built and handed over

source: green sparrow tattoo co. · toasted bean

monthly platform fees charged to keep a delivered system running

source: the ownership rule

approvals delegated to an AI in a client workflow

source: the human-in-the-loop rule, by design

Kept general

Not everything I work on is client work, and not all of it can be described. Alongside this practice I am an AI program manager at Mercedes-Benz North America, on the government and defense team, building internal AI tooling for business development inside their governance framework. That is a role rather than an engagement, so it is named here and not listed above. There is also a confidential enterprise recruiting assistant behind an NDA. I can talk about the shape of both and the standards they taught me, not the specifics, and that discretion is the same one your project gets.

If one of these looks like your problem, the conversation starts with your workflow.

Start a project