Portfolio
Selected Work
These are systems I've designed, built, and shipped: real work, real users, honestly described. Most were built at a large organization that I don't name, to protect its internal details. Where I don't have a measured number, I don't give one.
One System, Every Department
A shared system that turns scattered outreach and follow-up into one consistent, automatic process across departments.
Problem
Outreach and follow-up were spread across several internal groups. Coverage depended on who was handling it that week, and there was no shared way to draft on-brand replies or keep track of what had been sent.
Approach
I built a custom AI assistant trained on the team's outreach playbook. It uses an instruction-based setup, so nobody needs a paid subscription to use it. It draws on the team's own approved messaging, tone guidelines, and links. I tested it on realistic sample scenarios before it touched a live conversation.
It connects to an automated routing system. Each logged interaction goes to the right group's tracking log, and a 7-day follow-up reminder goes out on its own. The system runs on the structured data the reply assistant in "On-Brand Replies, Every Time" already produces, so nobody has to re-enter anything.
Outcome
In use across several teams, built on the reply assistant in "On-Brand Replies, Every Time". It is new and has not been tested through the busiest weeks of the year, but it is real, deployed work.
ParkPals Crew, an iOS App
A free iPhone app that helps a group of friends find theme park days that actually work for everyone.
Problem
Planning a park day with friends means juggling who has which pass, who is free, and who still needs to reserve a spot. That usually turns into a long group chat and a lot of back and forth.
Approach
I designed and shipped the app myself, start to finish, using AI coding tools. People enter their pass, start a crew, and invite friends with a short code. A shared calendar then shows the days the whole crew can go and the days only some can. Members can mark busy days with private notes and choose who sees them, propose a day with a short message, respond to someone else's proposal, and track who still needs to book a reservation.
It went through Apple's App Store review and is published under my own developer account. It supports light and dark mode, and it is an independent app with no connection to any park.
Outcome
Live and free on the App Store: ParkPals Crew. It is the clearest example of the app work I offer, from first idea to a published build.
This Website
A fast, plain website for a new solo practice, built by hand and kept easy to maintain.
Problem
A new practice needed a clear, credible website, with a simple way for visitors to get in touch and for me to be notified, without extra platform fees or ongoing maintenance.
Approach
I built it by hand from plain pages with no framework, so it loads quickly and is easy to change. Updates go live automatically once approved. The contact form has labeled fields, a spam trap, and a skip link for keyboard users, and each submission sends me an email and a private alert. It uses no tracking or cookies, and a privacy page explains what is collected.
Outcome
Live at marginnmethod.com, and a working example of the website work I offer.
On-Brand Replies, Every Time
An AI assistant that keeps outreach replies consistent, on-brand, and fast to send.
Problem
Replies needed a consistent tone, approved messaging, and the right links every time. Writing each one by hand was slow and varied by who was handling the conversation, especially when volume spiked.
Approach
A custom AI assistant trained on the internal outreach playbook and approved messaging, with a reference list of links. It drafts replies in the right tone and adds the correct links automatically. I tested it on realistic sample conversations first, checking tone, accuracy, and links before anyone on staff relied on it.
It also produces a short structured summary of each interaction: a fit rating, a category, and a partner or industry tag. Every conversation becomes usable data instead of a one-off email. That output is what makes the routing in "One System, Every Department" possible.
Outcome
Drafts the team's outreach replies today. The data it produces feeds the routing, logging, and 7-day reminders in "One System, Every Department".
One Dashboard, Every Angle
A manual reporting task, replaced by a live dashboard that updates itself every day.
Problem
A client-facing team received interaction data as several emailed reports a day, each holding part of the picture. Turning them into something usable meant collecting, combining, and reformatting every report by hand. That added up to hours a week for one person and far more across the team over a year.
Approach
I built a script that watches the inbox for those reports, then reads and combines them into one live data source that feeds a dashboard. Anyone on the team can open it when they need it, without waiting for someone to compile it. I tested it on weeks of realistic sample reports first, checking for reading errors and formatting problems, before it touched a real inbox.
The dashboard is filterable and refreshes automatically every day. It tracks trends, surges, and baselines across categories, groups, and staff type.
Outcome
A live dashboard with an executive overview, a staff-workload view, and a demographic view for different audiences. It is used across the organization, from staff to senior leadership, for performance reviews, year-end reporting, staffing decisions, and checking that service reaches everyone equitably. All figures are aggregated and anonymized before anyone sees them.
Never Miss a Renewal Again
A safety net that catches expiring agreements before anyone has to remember to check.
Problem
A partnerships team tracked over 200 active agreements by hand in a shared file. Catching expirations depended on someone remembering to check the list in time, with no backup if that person got busy or moved on.
Approach
An automation connected to the team's existing tracking file that watches every agreement's expiration date. It reads the file the team already used, so there is nothing new to learn or keep in sync. I tested it on a copy of the real file first, checking the reminder timing before it sent a single real reminder.
It sends reminder emails at 90, 30, 14, and 3 days before each agreement expires, to everyone who needs to act on it. Coverage no longer depends on one person's memory or schedule.
Outcome
Built and ready, and currently in requirements-gathering with stakeholders ahead of a full rollout, so the recipients and timing match how the team really works.
One Home for Every Resource
A single, always-current hub that replaced scattered documents with one place to browse and copy from.
Problem
Team members wanted quick access to ready-to-use client-facing materials, such as templates and standard wording. Finding them meant digging through scattered documents, shared folders, and old emails every time.
Approach
I built an interactive hub of over 80 click-to-copy resources, organized by category, using AI coding tools and several coding agents working at once. I built, tested, and refined it step by step, checking layout, search, and copying against real usage and staff feedback before it went live.
It sits inside the organization's existing internal site, so there is no new login or link to learn. Anyone who could already reach the site could use it right away, with no training.
Outcome
Adopted organization-wide since launch, well beyond the team it was built for. Leadership is working to standardize it as a shared resource.