I've spent three decades writing software, from the bare metal to the boardroom. I wrote the original shopping cart for MotherNature.com in the mid-90s, led Evidence.com, the police body-camera video platform, at TASER (now Axon), and guided engineering teams through fast growth and public listings. Today I build fast Rust and Go systems and open-source tools, explore deterministic AI, and lead engineering at Estrin Group.
From a teenage founder in the 90s to CTO of public and private companies, and back to writing code every day.
IPO (1999) • $600M Peak Market Cap
I co-founded the company at 15 and was the original CTO. Store software was still young, so I wrote the company's first shopping cart myself. It grew into the world's largest online vitamin and supplement store, and went public in 1999. After the IPO, I led the team that built version 2 of the site.
Public Company • $73M Peak Market Cap
I co-founded this media company at 19 and took it public. It produced four cable TV shows and earned revenue from broadcast, streaming, DVD and international content sales. I ran the websites and infrastructure behind the shows, and led MediaLimiter, the company's anti-piracy software.
Technology Firm • Santa Monica
I founded and ran this technology firm. Its main product, the Odin Organic Framework, was an enterprise CMS that clients licensed by the month. It also built Odin Assemble, a framework for building web apps quickly, and Avalerion, a network of sites with contextual advertising. Beside the products, the firm offered SEO, code reviews, security audits and custom software.
Public Company • TASER International
I led development of Evidence.com, the SaaS platform for police video evidence, with 9 direct reports and 20 more across the division. I also led the ETM (Evidence Transfer Manager), a device that moves video off officers' cameras securely and keeps the chain of custody intact. I was a speaker at the 2009 TASER Tech Summit.
MindTouch • Chatmeter • MojoPages • Geekli.st • oNoise
At MindTouch, I was SVP of Engineering for its knowledge base product and Zendesk integration. Then I served as CTO, full-time or fractional, at four startups: a local brand management platform, an internet yellow pages site, a developer community, and an audio collaboration app. Each one needed something different: hiring, Scrum, architecture, fundraising, and plenty of hands-on code.
iPad Uniform Ordering App
I led delivery of an iPad app that replaced printed catalogs. Restaurant buyers dressed rotatable 3D avatars in Chef Works garments, previewed them over a photo of their own restaurant, and turned the result into an order.
TapWise • IoT Draft Dispensing
As the top technology executive, I reported to the CEO and managed software, hardware, R&D and QA. TapWise put cloud-connected controls on draft taps: portion control, tap locking and volume analytics, at venues like Nationals Park and the Kennedy Center. I also led a battery design that let TapWise move from fixed bars to mobile venues.
Vertically Integrated CBD • ~380 Acres
With nearly 380 acres in Ventura County, TheraJoy once held the record for the largest outdoor CBD grow in California. It grew its own hemp, extracted it and made its own products. I built METIS, the system that tracked every crop from the field to the finished product, in PHP and Laravel. I also built the online store and worked on branding and packaging.
Software & AI Engineering
I have consulted since 2013, and that practice is now Estrin Group. I build products and AI systems for clients and serve as a fractional CTO. In one recent rebuild, a client's data pipeline went from finding the data it needed 9% of the time to 81%.
Libraries, CLIs and APIs, built out of curiosity and real need.
One AI reviewer misses things and sometimes invents problems. ATCR sends a code change to a panel of reviewers on different models and providers, each with its own focus. A Go reconciler then groups their findings, removes duplicates, and scores each one by how many independent models agree. By default, findings the panel does not confirm go to a separate file, so they are kept but do not crowd the report. It runs as a CLI, an MCP server or an Agent Skill, and works with local models, so no code has to leave your machine.
An npm package that gives Node.js apps one simple interface to 36 AI providers and hundreds of models, so switching providers is a small change, not a rewrite. It handles chat, streaming and embeddings, with failover, response caching, retries, and repair of broken JSON.
An API that measures how a person writes. I built it first as a Python proof of concept. Once the features were proven, I rebuilt it in Rust for speed, then spent that headroom on features that would cost too much in Python: embeddings, BERT content-safety checks and optional GPU speed-up. It powers my revoice skill, which rewrites text so it sounds like you, not like an AI.
What three decades of launches, outages and changing tools have taught me.
If code can do a job reliably, code does it. I give AI only the small parts that need judgment, and I check its answer with code before the next step runs. The work gets cheaper, faster and easier to trust.
I pick the language by the problem, not by habit. Years of Perl, PHP, C# (Mono), TypeScript/Node, Python, Go and Rust taught me that every language is good at something:
As a CTO, my job was to hire well and delegate. As a consultant, I work one on one with each client, and there the person who designs the system should also build it. Nothing gets lost in a hand-off, and the engineer who made each choice is the one who answers for it.
I run private AI services in my own lab, so code and data stay on hardware I control. When a task does need a cloud model, I first strip personal data where it matters, using a PII filter such as Microsoft Presidio, configured for the job. Sending data out is a choice, not the default.
Client work, open source, the occasional founder conversation, and the lab that runs it all.
I lead engineering at Estrin Group: product builds, AI systems and fractional CTO work for clients.
ATCR, LLM Interface, and a growing set of Go MCP servers and CLI tools.
A sounding board for technical founders, from someone who has been both the founder and the CTO.
17 machines running private AI services, a code review pipeline and hosting, backed by UPS zones, a whole-house battery and solar.
For product builds, AI systems, hosting or fractional CTO work, start at Estrin Group.
Want to talk about systems, deterministic AI, open source or 90s web history? Send a note. I read every message myself.
For client work, start at Estrin Group ↗.
