Read the message
Start with what the resident actually said. Do not flatten a broken lock, a rent question, and a noisy hallway into the same score.
Co-founder at ResiDesk
Right now, that means building ResiDesk with property teams. Residents text about rent, repairs, renewals, parking, and everything else that can make a building hard to live in. We help the team answer well, remember the context, and fix the things that keep coming back.
Before ResiDesk, I built software for financial advisors at BlackRock. Then I ran product, engineering, and data at Climb Credit. Both jobs taught me the same thing: sit with the person doing the work, measure what changed, and keep fixing the parts that fall apart on a normal Tuesday.
ResiDesk
A resident might ask about a broken lock in a text, mention it again on a call, and leave a review after the third attempt. Property teams hear versions of that story all day across texts, tickets, surveys, calls, and renewal notes. Answering one message is useful. Knowing the repair is still open is the part that matters.
We started ResiDesk after seeing how much of this work lived in separate inboxes, property systems, spreadsheets, and people's heads. I co-founded the company, and my week moves between product and data, customer calls, sales, hiring, and helping the team get past whatever is stuck.
AI helps us read and connect more of that history. It still has to know the building, follow the policy, admit what it does not know, and hand the work to a person when the person needs to decide.
See ResiDeskStart with what the resident actually said. Do not flatten a broken lock, a rent question, and a noisy hallway into the same score.
Find the lease, building policy, earlier messages, open work order, and what the team has already tried.
Answer the question, open the repair, explain the policy, or ask someone to decide. Make it clear who owns the next step.
If move-in instructions confuse twenty residents, the answer is not twenty better apologies. Fix the instructions.
The road here
I studied applied physics at Cornell and spent time in an electron microscopy lab. Part of a magnetic-noise experiment meant doing the same slow analysis again and again. I wrote a small program to do it faster, and a researcher got hours back. That was the first time software felt useful to me.
The code was not ambitious. It removed a tedious part of someone else's day. I still start there.
At BlackRock, I built analytics software for financial advisors and the people planning their retirements. I won the internal Aladdin Hackathon three times. A prototype could still win the room and lose the next morning to a spreadsheet that people knew how to use.
At Climb Credit, I ran product, engineering, and data while we built loans for career-focused education. We brought completion rates, graduate earnings, and employment outcomes into the product instead of pretending a credit score told the whole story.
A resident's message can become an answer, a repair, a renewal conversation, or a decision about how a building is run. I want our product to follow that work far enough to know whether anything got better.
Homes covered in a four-month study of what residents said about Wi-Fi and how it related to renewals. Read the study
Climb's annual loan volume grew from roughly $1 million to $300 million while I led product, engineering, and data. The early Climb story
Annual recurring revenue reached by the advisor analytics product I helped build from its first version at BlackRock.
How I work
Ask what they are trying to finish, what information they are missing, and what happens when the queue gets busy.
I want the person who found a problem to feel glad they raised it. Then we can count the cases, find the cause, and fix the part we control.
The real test comes when the queue is full, the policy is unclear, and nobody who built the demo is in the room.
If a person has to correct the same mistake twice, I want the second correction to become part of how the work gets done next time.
Notes
I usually start because I caught myself saying something too neat in a meeting. Writing forces me to find the example, check the claim, and decide what I would actually do.
With AI, I care less about which model won this week than whether the product can read the right information, use the right tools, show what it did, and stop when it is unsure.
I am also starting to look more seriously at fair housing as AI takes on more leasing and resident work. If software helps decide who gets an answer, which problem gets attention, or when a person steps in, we should be able to test whether it treats people fairly.
Read what I've been writingConversations
Three videosThings I've written or joined
9 piecesModels keep getting better and cheaper. The hard part is still giving one the right information, tools, checks, and a clear place to hand the work back to a person.
2026 AI from the owner and operator seatA BuiltWorlds conversation about the questions owners ask before they trust AI with underwriting, property work, or a decision that moves money.
2026 What are residents trying to tell us?Dom Beveridge and I talked about resident comments, the decisions hiding inside them, and why a sentiment score is not the end of the work.
2025 Is bad Wi-Fi putting renewals at risk?A four-month study covering more than 30,000 homes across eleven states, built from what residents actually said about their internet.
2025 A good chatbot is still not enoughOwners care about collections, renewals, repairs, and why someone may move out. A fast reply matters only if it helps with the rest of that work.
2024 What resident conversations can changeLaw360 wrote about how ResiDesk helps property teams hear residents sooner and carry what they learn into retention and day-to-day operations.
2024 Putting AI inside real estate workA longer interview about why answering a resident is different from finishing the repair, renewal, or follow-up the answer creates.
2024 Practical AIA profile of my work at BlackRock, Climb, and ResiDesk, and the habit that followed me through all three: start with the person doing the job.
2016 Climb Credit in TechCrunchThe early Climb bet was simple to say and hard to build: finance education based on whether the program helped graduates find work and earn more.
Contact
I am always glad to hear from founders, housing people, and teams trying to make AI useful in a real business. I also enjoy helping with product choices, early company problems, and the ordinary mess that does not fit neatly into a job description.
You do not need a polished pitch. A few honest sentences about what happened, what you have tried, and where you are stuck are plenty.
I am especially useful when there is a customer to understand, a product decision to make, or a team that needs a clearer next step.