On the work that compounds.

I've been writing software for eleven years. What I do for a living is build the parts of large systems that other people end up standing on. End to end. React the user touches, Java services behind it, AWS underneath. Not full stack the way a frontend engineer says it. Full stack the way someone who has shipped React, Java, DynamoDB, and a Spark job in the same project, the same week, with no one in the middle.

Scale is the fun part on the whiteboard. The throughput numbers feel like cricket scores. The trouble starts later, once the easy answers run out and you're staring down consistency models, partial failures, p99 latency, and a bug that only shows up when three things are slightly off at once. That's where I live.

Amazon

Prime Video first. I owned growth experiments on Freevee end to end. What the viewer saw on screen, the Java services serving it, the Kinesis and Flink pipeline streaming ad watch-events, the Glue and Redshift platform behind NextUp's recommendations over IMDb's sixty-five million records, and the twenty-two A/B experiments running on top. The whole stack was mine. The work sat behind forty-four million dollars in measured streaming revenue. The number isn't the point. The point is it shipped.

Now I'm at AWS, on the systems AWS itself runs on. The topology graph that knows where every rack lives, the data layer that stores it, the services that read and mutate it, and the scaling that keeps all of it honest at the rate AWS grows. Closer to the metal than streaming. The consequences of a wrong setting are more interesting.

On the side of all that, I lead APEX, a Claude-based multi-agent system that resolves recurring datacenter ops tickets end to end without humans in the loop. APEX is the most interesting work I've done. It's also the work I've had to be most careful with. Confident-sounding agents in front of real infrastructure is a category that deserves more thought than it usually gets.

Before Amazon

Databuoy. Public-safety IoT. I was the only software engineer on ShotPoint, a system that used arrays of acoustic sensors to localize gunshots in 3D in real time. I built every layer. The Java microservices. The heartbeat platform scaling to twenty million sensor records. The calibration tooling. The React and Three.js visualization that put the real-time location of an active shooter into a responding officer's hand.

Deployed where seconds matter. There is a real chance some of what I wrote helped get someone home alive. I still think about that work when I'm asked what kind of engineer I want to be.

The thread

What I'd like the things I build to do is compound. For other people. Infrastructure others quietly rely on. Systems that keep someone safer. Platforms that pay back the trust placed in them. That's why streaming advertising mattered to me the way it did. It's why the public-safety work was easy to care about. It's why I like that most of what I do at AWS now is invisible to the people who benefit from it.

The distance between a system you can sketch on a napkin and a system you have to run at three in the morning when a partition you didn't think was possible is happening anyway. Closing that distance is most of what senior engineering is.

Outside the work

I moved to Denver for the mountains and stayed. Twenty-two of the sixty-three American national parks down, forty-one to go, at the pace of someone who isn't in a hurry. I photograph the kind of light you only get on a ridge in late afternoon. Badly, but improving.

I keep an open ear and an open calendar for the arts. Musicals, concerts, exhibitions when they pass through town. I don't chase every show. I let the right ones find me, and I show up when they do.

On music I am a polytheist. Old Hindi songs from the seventies for slow days. Linkin Park for the rest. Coldplay either way. I'll fight anyone who says these don't belong on the same playlist.

I invest with patience and concentration. Most of what I own I expect to hold for years, sometimes decades. I read business and economic history, accounting, behavioral psychology, and enough mathematics to follow the arguments. I write up the companies I follow before I take a position. So I can read my own thinking back to myself later.

I box. Two or three times a week. The discipline of staying loose under pressure, breathing through a hard round, and reading another person in real time has a strange amount in common with debugging a system that is failing in distribution. Both ask the same question: stay calm, what's actually happening?

And I read widely. Psychology, philosophy, economics, science, tech. The questions I care most about live at the seams between those fields, where one discipline borrows from another to explain what neither could explain alone.

That's most of it. If something here resonated, you know where to find me.