About
Who I am
Software engineer, currently between SF and SoCal. I’ve built and shipped at Tesla, Labelbox, First Round Capital, and UW. These days I build agents and the infrastructure they run on, from bare metal to browser — 115 and Lumbridge, a neocloud whose compute layer runs on unified-memory NVIDIA hardware. Most of what I learn ends up running on my own hardware as something you can actually use: a recording studio, a watch-together theater, an arcade, a 3D brain viewer.
Philosophy
Timeline
AI Engineer World's Fair — PodMan
2nd place out of 95+ teams at the AI Engineer World's Fair Hackathon, plus the LiveKit and DigitalOcean track awards. Built PodMan — an ambient voice-AI pair-programmer that watches a team's shared room and flags duplicate work and merge collisions before they cost anyone time. Shipped to production over a weekend.
Lumbridge
A neocloud you can walk around — your whole cloud (database, hosting, compute, AI agents) set up for you as a 2D world you move through, rather than five vendors and five bills. Underneath it's a cloud reseller packaged like Vercel: sign into World 1, rent a plot, get your own provisioned world. Its compute layer runs on unified-memory NVIDIA hardware, where CPU and GPU share one pool and over-committing wedges the whole machine before the OOM killer gets a chance — so the node agent admits work against a hard memory budget and supervises every model it starts. Early, but running end to end.
The karti.ai network
A set of sites I run on my own hardware: a browser recording studio, synced watch-together theaters, a multiplayer arcade, and a GPU volume renderer for MRI data. Most of them exist because I wanted to learn something specific and a real user-facing thing is the honest way to find out whether it works.
115
A personalized trading harness — software that wraps one human trader in a full engineering stack, so a single trader or a multi-node desk can run a real book around the clock. Signal to fill is all software I own, with a human in the loop on the calls that matter. Pre-alpha research preview.
UW
Software engineer building tools for retail traders. Real-time data pipelines, option flow analytics, and systems that handle millions of events per day.
First Round Capital
Startup Solutions Engineer — helped portfolio companies ship faster. Internal tooling, automation, and the infrastructure glue that keeps early-stage teams moving.
Labelbox
Solutions Engineer at the data labeling platform. Built integrations, solved customer problems, and learned that good software starts with understanding the user.
UC Davis
Computer Science & Managerial Economics. The mix of engineering and business thinking shaped how I approach problems — build the right thing, not just build things right.
Tesla
Business Development in Hawthorne. Learned how hardware and software come together at scale. Watched a startup become the most important car company in the world.
VESC
Engineering intern at an electric skateboard startup. Worked on open-source motor controllers (VESC). Company got acquired by Boosted Boards in 2012.