Outcomes over output
I measure a product by what changes for the business and the people using it, not by how much ships.
Proof My lead-scoring product became Procare's main source of net-new pipeline.
Johnathan Margheret
Six years turning ambiguous, regulated problems into products people actually use, across B2B SaaS and enterprise data platforms.
I build complex data and AI products that turn strategic ambiguity into shipped solutions with measurable business impact. Over 6+ years spanning B2B SaaS, internal platform development, and enterprise data infrastructure, I've shipped AI document-extraction applications, machine learning platforms, and analytics products that together have driven millions in revenue and cost savings.
I bridge executive leadership, engineering, and business stakeholders in complex, regulated environments. I captain end to end, from product vision and discovery through delivery, launch, iteration, and adoption. Today I lead data product strategy at TEAM Services Group, owning a modern stack spanning Snowflake, dbt, Dagster, Power BI, and Streamlit across 120+ integrated systems. When I'm not shipping enterprise data products, I'm still in the garage honing my skills and building apps that solve problems.
I measure a product by what changes for the business and the people using it, not by how much ships.
Proof My lead-scoring product became Procare's main source of net-new pipeline.
I don't back off hard problems, whether it's a messy platform migration, a heavy lift at the gym, or a steep line on the mountain.
Proof Unified 16 organizations onto one data platform, 0 to 1 in under six months.
I learn by building. When a new tool or idea shows up, I want to know how it works and what it makes possible.
Proof After hours I built WanderMate, an outdoor trip decision engine with transparent scoring, using Cursor and Claude Code. It's the updated version of my first trip planner, TripForge.
If a process can be faster, cleaner, or automated, I'll find it, and then measure whether it actually got better.
Proof Rolled out Agile/Scrum that increased team delivery velocity by 300%.
I grew up playing soccer and still play. I lead the same way at work: aligning executives, engineers, and the business on one plan, and putting the team first.
Proof I lead a cross-functional data team and an outcome-driven roadmap set with executive leadership.
TEAM Services Group
National in-home healthcare platform · 15+ brands · 50 states
Procare Solutions
#1 US childcare management software · 37,000+ customers
Seamless.AI
AI sales intelligence platform · LinkedIn Top 50 Startups
AI document-extraction application for in-home healthcare at TEAM Services Group.
Problem High-volume documents were keyed in by hand across 15+ brands. It was slow, repetitive, and easy to get wrong.
What I built A Python and Streamlit application that uses an LLM to read each document and extract the fields that matter, with a human-in-the-loop review step so reviewers approve instead of transcribe, and an integration that writes clean data downstream into the EMR. Built first for compliance documents, then extended on the same pipeline to a second workflow.
Outcome Millions saved each year, with manual entry and reconciliation work nearly eliminated.
Machine learning lead generation platform built for Procare Solutions.
Problem Sales had no dependable source of net-new pipeline. Reps picked leads on gut feel and a lot of them went nowhere.
What I built An end-to-end lead-scoring product: a machine learning model defined with data science, plus a Salesforce integration that puts each score in front of reps inside the workflow they already use.
Outcome It became the company's main source of net-new pipeline and lowered acquisition costs.
Customer health product for Procare's Tuition Express payment platform.
Problem The business had no live read on product health, so nobody saw which payment accounts were quietly slipping until it was too late.
What I built A customer-health product in Power BI fed by Pendo product signals. The harder part was deciding which engagement and usage metrics actually meant an account was at risk, rather than just charting everything.
Outcome Gave the team its first real-time pulse on product health and the accounts worth worrying about.
Predictive churn and account-health scoring for Procare's 37,000 customers.
Problem Cancellations kept catching the team off guard. There was no early warning across a base of 37,000 accounts.
What I built A predictive churn model and account-health score, with features designed alongside data science and a playbook built with customer success for what to do when an account lit up.
Outcome High-risk accounts got flagged before they cancelled, protecting recurring revenue.
An iOS app that scans your photo library and organizes it into albums by moments, trips, and categories, with search that actually works: by category, date range, or plain language. Claude vision handles tagging, captioning, and quality scoring, and embeddings power semantic search. Albums are virtual, so your Apple Photos library is never modified unless you export, and every action can be undone.
An outdoor trip decision engine for campers, hikers, and anglers across Colorado, Wyoming, and Utah. Enter your location, skill level, vehicle type, and crowd tolerance to get ranked, explained recommendations drawn from 24 curated trip zones. A transparent 0 to 100 scoring algorithm weighs activity fit, drive time, off-grid capability, and eight other dimensions, so you know exactly why a trip made the list.
An original bullet-heaven game set in 2167, where a sentient AI collective called NEXUS has turned on the last human settlements. You play a Guardian: move with one thumb while weapons auto-fire, survive escalating machine waves, and destroy each sector's Signal Core. I ran it like a product, with an approved design doc, a milestone plan that proves "is it fun?" with one level first, and a 42-issue backlog.
Consistent training. Small gains, every week. The gym is the one place the roadmap doesn't follow me.
Not the groomed stuff. Cliffs, steep chutes, and whatever the mountain hides off the map.
Grew up playing and still do. I get inexplicably competitive about it.
Open to conversations about data and AI products: roles, advising, or just comparing notes.
Denver, CO · Remote-friendly