v3.0 Now with applied AI

A CTO who ships.

Steven Ou is a three-time founding CTO and self-taught builder. Scales consumer and enterprise platforms from zero to nine-figure revenue. Applies AI to products people pay for.

Read the docs↓ See case studies
No consulting-speak No slide-deck theater Just ships
Brands using what I've built at Cavalry
Harry's Touch of Modern Naadam Dolls Kill Geologie Cameo
Cavalry backed by
SV Angel Bling Capital
Capabilities

What I do best.

Four capabilities, all backed by twenty-six years of shipping real software to real users.

01

Build from zero.

Stand up full product stacks from nothing — web, mobile, backend, data, internal tools. No legacy crutches. No incumbent playbook.

web · ios · android · infra · data
02

Ship fast.

Self-taught programmer since age eleven. Still hands-on. Still reads pull requests. The speed compounds across a career; the quality doesn't slip.

shipping since 1999
03

Turn AI into product.

I take what works and ship it into production — with the security, integrations, and reliability enterprise buyers actually need. That's the work I've spent the last four years on.

llm orchestration · deterministic policy engines · production systems
04

Operate technology as a business.

A founder's P&L discipline, a platform-builder's architectural depth. Technology as the engine of the business, not a line item beneath it.

p&l ownership · board engagement · security & compliance · org design
Compatibility

System requirements: met.

Philosophy
Developer velocity over everything.
Languages
Ruby · JavaScript · TypeScript
Frameworks
Rails · Hotwire · Stimulus · Sidekiq
Data
PostgreSQL · Snowflake · Valkey
Infrastructure
AWS (ECS, RDS, S3, CodeDeploy)
CI/CD
GitHub · GitHub Actions
AI
OpenAI · Anthropic · Gemini
Compliance
SOC 2
Case studies

Three deployments. Same operator.

Past performance. Because future promises are for pitch decks.

Case 03 · Current
2021 — Present
Cavalry AI
Co-Founder · CTO
60–80%
Autonomous resolution
78%
Naadam peak Q4
360%
Harry's ROI on CX
25×
Faster resolution
Problem
Enterprise e-commerce teams have tried DIY AI tooling and failed. Workflow-based chatbots top out around 20% resolution. Buyers have lost faith that AI support can actually work.
Approach
Pairs stochastic LLMs with a deterministic OS layer — a policy engine that enforces eligibility and executes every resolution in the real systems of record. The LLM handles understanding and language; it cannot act outside the policy. That split is how we drive accuracy and reliability above what pure-LLM or pure-workflow approaches can reach. Integrates with the systems e-commerce teams already run on — Shopify, BigCommerce, Zendesk, Gorgias, Klaviyo, Recharge, and more. White-glove onboarding takes real work — weeks of careful configuration, because that's what produces results the brand can stand behind.
Outcome
Agents resolve 60–80% of complex support cases autonomously — not the 20% workflow-tool baseline. Shipping to brands including Harry's, Naadam, Dolls Kill, Laundry Sauce, Geologie, and Touch of Modern.
Case 02 · Exited
2012 — 2020
Touch of Modern
Co-Founder · CTO
$200M+
Annual revenue
17M+
Registered users
100+
Employees
$17M
Capital raised
Problem
Build an e-commerce platform that could merchandise 250+ new products daily, scale to millions of users, and defend margin against Amazon-class competitors. No legacy stack, no incumbent playbook.
Approach
Built the complete technology organization from zero: web, native iOS and Android apps, merchandising, fulfillment and logistics software, data and analytics infrastructure, and internal operational tooling. Positioned technology as the competitive moat — kept nearly every business function in-house.
Outcome
Scaled to $200M+ revenue and 17M+ users. Inc. 5000 two consecutive years, Inc. 30 Under 30 Rising Stars, and Forbes 30 Under 30 for the co-founding team.
Case 01 · Foundation
2010 — 2012
Skyara & RAVN
Co-Founder · CTO
i/o ventures
Incubator
2
Companies, one team
Day 1
Started at graduation
UBS
Offer declined
Problem
How do you learn what it actually takes to build a real company? Answer: try, fail honestly, carry the lessons forward.
Approach
Turned down a full-time UBS offer the day of UPenn commencement. Co-founded Skyara (local services), then RAVN. Selected into I/O Ventures incubator.
Outcome
Proved the unit economics in that category wouldn't scale. Pivoted the team into what became Touch of Modern. The full-stack foundation and the founding partnership carried forward.
Changelog

Shipped recently.

A working career, in reverse chronological order.

v3.0.0
2021 — now
Cavalry AI. Shipped the OS layer — a deterministic policy engine paired with stochastic LLMs to deliver autonomous customer support at 60–80% resolution, well above what pure-LLM or pure-workflow systems reach.
v2.0.0
2012 — 2020
Touch of Modern. Built the full platform: web, iOS, Android, merchandising, fulfillment, data. Scaled to $200M+ in revenue and 17M+ users. Shipped every day for eight years.
v1.0.0
2010 — 2012
Skyara & RAVN. Founded two companies out of I/O Ventures. Learned what scales and what doesn't. Carried the founding team forward.
v0.1.0
2006 — 2010
University of Pennsylvania, B.A. Economics. Summer research at ITG (quant trading), capital markets at Chinatrust, systems development at CineNow Singapore. Accepted UBS offer; turned it down to start Skyara.
v0.0.1
1999 · Age 11
Initial commit. First line of code. Self-taught. Never stopped.
Roles

Senior, two ways.

Current
Exec
Venture CTO
For companies building real product, where the CTO role calls for a builder.
  • Full ownership of technology, platform, and applied AI
  • Founder's P&L sense, platform-builder's depth
  • Scope across engineering, product, and architecture
LinkedIn
IC
Principal / Staff Engineer
For companies that need the builder without the exec layer.
  • Applied AI, systems design, end-to-end product engineering
  • Architecture, shipping, and hands-on ownership
  • Deep technical fluency on real production problems
LinkedIn