SELECTED WORK / 2021—2026

Open source · AI products · agent infrastructure

Systems shipped at the frontier of AI.

I build the layers that make intelligence useful: generative interfaces, agent infrastructure, databases, product loops, and the operating systems around them. This is selected work from teams I have joined, led, and founded.

2.05Mcombined social impressions across CopilotKit launches

1.5KGitHub stars for OpenGenerativeUI

723GitHub stars for OpenTag

1 dayto solo-build the first OpenGenerativeUI release

02 / Founder build

Folderbase.ai · Database core + cloud platform

A database for agents that still feels like a folder.

Folderbase turns ordinary folders into structured, versioned workspaces that humans and agents can understand across sessions—without hiding files behind a proprietary format.

CoreApache-licensed Rust library, CLI, protocol, templates, migrations, versions, and sync primitives.

CloudNative app, managed synchronization, sharing, recovery, web workspace, and version-pinned agent Change Sets.

PrincipleFolders stay folders. Agent context, identity, policy, and history become durable infrastructure.

01Ordinary files

Native formats · Native tools

02Folderbase core

Identity · Policy · Versions

03Cloud platform

Sync · Sharing · Agent work

03 / Ecosystem collaboration

Google A2UI · CopilotKit

Giving a UI protocol a place to be composed.

I built on the A2UI ecosystem through CopilotKit, helping turn a protocol into an explorable Composer with creation, workbench, gallery, catalogs, icons, and JSONL playback surfaces.

The work became part of the product story featured at Google Cloud Next 2026—connecting protocol-level thinking to an interface builders could immediately understand.

Open A2UI Composer

04 / Founder build

Chalk Agents

Infrastructure for agents doing real work.

Chalk is where I turn hard-earned agent engineering lessons into tools: from a testing browser built for coding agents to coordination, permissions, and governance for agent companies.

1 weekto build the first agent testing browser

3 months → 1 weekenterprise security project delivery

2first enterprise clients won

Explore Chalk

05 / And more

Years of product work.
Measured in outcomes.

Selected products and systems built across AI, venture studios, product leadership, and enterprise transformation.

0161,151

InterviewRoom.ai

Applicants served across 15,332 automated interviews; time-to-hire moved from 45 to 18 days.

022,000

SenatorMatch

Users in the first 48 hours for a civic-tech product built to make candidate research easier.

03200K+

bNeXt

Hours of annual work automated for an enterprise finance workflow; winner of SAP Hack2Build.

04100+

EngagePath / AI Coach

Prompt prototypes used to turn 16,923 profiles and 652 ICP patterns into product evidence in one day, not two weeks.

0519.06%

TripGuru

Year-over-year profit increase while support response time moved from two minutes to ten seconds.

061,292

MultiplAI Agents

Users in the first 30 days for an AI-agent product that generated $7,788 in early revenue.

Fractional leadership · advisory · selected builds

Bring a builder’s point of view to your problem.

I work with ambitious teams where product judgment, AI engineering, and organizational change have to move together.