Building ResiDesk

I build software for the messy work around customers.

Most of my time goes into ResiDesk, where product, data, and customers all meet. We help property teams answer residents, understand what keeps going wrong, and get each issue to the person who can do something about it.

Before ResiDesk, I built advisor software at BlackRock and later ran product and engineering at Climb Credit. The industries changed, but the habit did not: stay close to the customer, measure what changed, and keep going until the thing works on a normal Tuesday.

Arjun Kannan
I live in New York. Most days I am thinking about housing, software, or both.

Renting is expensive enough. Everything around it should be easier.

Residents tell property teams what is broken every day. It shows up in texts, reviews, tickets, surveys, calls, and renewal notes. Another inbox is rarely the answer. The team needs the history, the right policy, and someone who owns what happens next.

We started ResiDesk to make this work easier. I spend most of my time on data and product, and the rest on customers, sales, hiring, and whatever else needs doing that day.

Visit ResiDesk
01

Hear what happened

Start with the resident's own words. The score comes later.

02

Bring the context

Bring in the lease, policy, unit, earlier messages, and what the team has already tried.

03

Decide what happens next

Answer, repair, escalate, explain, or change the policy. Someone still has to own the decision.

04

Learn what repeats

When the same problem keeps coming back, show the pattern to the people who can fix it.

The industries changed. The way I work did not.

I grew up around research and studied applied physics at Cornell. Software became real to me in an electron microscopy lab. A small tool I wrote saved hours on a magnetic-noise experiment. That was when software stopped feeling like coursework and became a way to help.

Physics

Measure first. Then explain.

The first useful program I wrote saved a researcher hours. I still use that standard: did the work get easier?

BlackRock

Real stakes make the details matter.

At BlackRock, I worked across product and engineering on advisor tools. A prototype could win the room and still lose to the spreadsheet people trusted the next morning.

Climb Credit

Change the outcome. Then build it into the product.

As CTO and CPO, I helped bring graduate earnings into the product, data, and underwriting instead of treating a borrower's credit score as the entire story.

ResiDesk

Stay close enough to the work to know what to build.

At ResiDesk, that means listening to messy conversations, following the work those conversations create, and learning from problems that repeat every day.

7%

One ResiDesk program reported this lift after a property team acted on resident feedback sooner. Law360

$1M → $300M

Climb Credit's annual loan volume grew across this range while outcomes moved into the product. TechCrunch

$40M ARR

I helped take an advisor analytics product from zero to this run rate in my first year at BlackRock.

Show me the work, the stakes, and the person who has to live with the decision.

01

Start with what someone is trying to finish.

Until you know that, the model debate is usually a distraction.

02

Be direct without making people feel small.

Take the pressure seriously. Separate what happened from why it happened. Then make the next move clear.

03

A demo is not adoption.

I care about what happens when the queue is full, the edge case is real, and nobody is around to rescue the workflow.

I write to figure out what I actually think.

I write when an idea still feels fuzzy. The test is simple: did this help someone do the work, or did it just make the demo easier to sell?

Read the Substack

The best conversations start with the actual problem.

I am most useful when the problem involves housing operations, AI in real work, product judgment, or the practical work of building a company.

Good reasons to write: you have a customer problem, a messy rollout, or a decision that gets clearer with the right context.

Less useful: a broad AI inspiration call with no customer, no real work, and no next decision.