01
Know what you are trying to learn.
Start with the problem and the biggest unknown. Build just enough to get a real answer.
CTO AT 25MADISON ↗
I build AI products and the teams behind them.From the first idea to production.
I work with founders to figure out what to build, how to build it, and who we need on the team.
See the work/ 01 — THE WORK
The hard part is rarely the technology. It is getting the whole system to move together.
I work across the product, the software, and the team so a decision in one does not create a mess for the others.
02
The demo is the starting line.
A demo tells you the idea might work. Then you have to make it useful, reliable, and ready for real customers.
03
Replace magic with mechanism.
With LLMs, I want to know what works, what it costs, where it fails, and where a person still needs to make the call.
/ 02 — CAREER SPINE
Five companies. A lot of building.
I have started companies, scaled products, hired teams, and stepped into the hard technical work when needed.
01
25madison
Chief Technology OfficerCurrent
Helped launch more than a dozen companies, building the first product, technical foundation, and engineering team.
02Reify Health
Director of Platform
Built platform systems for StudyTeam as it grew into a global product connecting clinical-trial sites and sponsors.
03EvidenceCare
Chief Technology Officer
Moved clinical decision support into the EHR, with APIs that fit the way clinicians already worked.
04SimplyAgree
VP of Engineering
Helped turn manual legal-closing work into software for packets, signatures, and closing binders.
05Satchel Health
Technical Founder · CEO · CPO
Started a post-acute telemedicine company and took it from the first product to deployments across the country.
/ 03 — SELECTED CASES
What we built and why.
Three examples of the problem, the call I made, and what happened next.
01
25madison · 25m Health
Build new companies without starting from scratch every time.
- Situation
- At a venture studio, every company starts at zero. The product changes, but identity, security, data, deployment, and hiring still need answers from day one.
- My role
- I am the CTO and a hands-on technical partner to founders and portfolio teams.
- Hard decision
- Standardize the boring parts without forcing every company into the same architecture. Reuse what saves time; keep product decisions with the team.
- System built
- A repeatable path from thesis to production: diligence, first architecture, shared platform pieces, AI-assisted development, and hiring the first engineering team.
- Outcome
- More than a dozen venture-backed companies used these foundations to move faster while keeping control of their own product and architecture.
02
EvidenceCare
Put clinical guidance inside the physician’s existing workflow.
- Situation
- Clinical guidance only helps if it shows up when the doctor needs it. A separate app meant another login and another workflow, so it often went unused.
- My role
- I was CTO, responsible for the clinical platform and the engineering team building it.
- Hard decision
- Treat interoperability as product work. Put the recommendation inside the EHR and give partners a clean API.
- System built
- An API-first clinical decision-support platform that delivered guidance without asking clinicians to leave the patient record.
- Outcome
- The platform expanded into admission criteria and revenue-cycle work, giving hospitals a direct operational and financial reason to use it.
03
Satchel Health
Bring the clinician to the patient between in-person visits.
- Situation
- Skilled-nursing patients left the hospital with serious needs, but a clinician might only be on site every few weeks. That gap led to avoidable transfers and readmissions.
- My role
- I was the technical founder, CEO, and later CPO, from the first product through fundraising and national deployment.
- Hard decision
- Work with the care model already in place. The product had to fit existing staff, devices, reimbursement, and clinical workflows.
- System built
- A telemedicine platform that combined live video, connected diagnostic devices, clinical workflows, and operational data.
- Outcome
- The product reached facilities across the country. Public reporting cited ROI within 30 days, up to $1M in annual cost avoidance, and more than $2M raised.
/ 04 — ON THE RECORD
Writing, interviews, and public work.
A few places I have written or talked about how I work.
WRITING · 25MADISON
Building Fast, Staying Human
My notes from a conversation with Zach Zaro about speed, engineering culture, AI-generated code, and taste.
PUBLIC COMMENTARY · SIFT
The 2026 Engineer
Why judgment matters more as AI makes it easier to produce software.
INTERVIEW · ENGINEERING HIRING
What I look for when I interview engineers
What I look for in engineers, and why a coding test is only part of the picture.
PRESS · FORBES
Building Satchel Health
The story of building Satchel Health and the patient who made the work real for me.
/ 05 — NOTES
What I am learning.
Detailed notes on venture engineering, coding agents, and building with LLMs.
/ 06 — PROJECTS
Things I'm building.
I learn systems by taking them apart, building my own, and sharing what I find.
OPEN-SOURCE FIELD GUIDE
The Grok Book
Grok was open-sourced, so I built a tool to take its coding harness apart. The Grok Book makes that work public so anyone can see how the harness is built.
Visit thegrokbook.com ↗AGENT CODING HARNESS
Hive
My own coding harness. A frontier model coordinates sidekick models across a graph of work that can branch, run in parallel, and come back together.
Soon to be releasedTECHNICAL BOOK
The LLM Book
A practical guide to LLMs: how they are trained, how inference works, how to fine-tune them, and how to run open models on your own machine.
Soon to be released/ 07 — CONNECT
Have something
hard to build?
I like working with founders when the problem is real and the right answer is not obvious yet.
hello@macy.dev