Open to new roles

Hi, I'm Chris Rollins.

AI Engineer · Product & Business Leader · Austin, TX

I build production-grade AI and full-stack systems - RAG pipelines, governed model-call layers, audit-grade data pipelines - and bring the product judgment of someone who has run the whole business, not just shipped the code.

About

A bit about me

I'm an AI engineer and technical product leader with 20 years of building systems and running the businesses around them. I started out building enterprise data and BI systems with large enterprise technology firms, then spent 15 years founding and running healthcare and construction ventures that served thousands of patients and generated over $9M in revenue, and now run an independent AI practice building production systems, automations, and data pipelines across legal, healthcare, construction, and SaaS.

My work is turning slow, manual, or seemingly intractable problems into reliable systems: RAG pipelines with governed, cost-instrumented model calls; responsible-AI products that keep a human in control of every write; and audit-grade data pipelines that reconcile fragmented records to the penny. The projects below are sanitized, self-contained versions of those production systems, rebuilt so you can read the architecture and the code without the proprietary parts.

I care about clean seams, honest failure handling, and the details that make a system trustworthy: tests that run offline, security enforced in CI, and provenance you can audit. Having carried whole businesses across sales, delivery, finance, and compliance, I build for the outcome, not just the ticket. For my own AI platform that meant building the growth engine as deliberately as the model: marketing automation, lead capture and segmentation, funnels, and conversion tracking, so it was built to find and keep customers, not just to demo well. What I want next is a place to put both halves to work - deep domain knowledge in healthcare and construction alongside hands-on AI engineering - helping a team turn AI into something their clients can actually rely on.

Skills & tools

AI / ML
RAGLLM orchestrationPrompt engineeringLLM-as-judge evalspgvectorResponsible AIAgentic workflows
Languages
TypeScriptPythonSQLJavaScript
Frontend
ReactNext.jsTailwind CSSZod
Backend & Data
Node.jsPostgreSQLSupabaseRow-Level SecurityREST / RPC APIsRedisSQLite
Infrastructure & Tooling
VercelGCPDockerGitHub ActionsVitest / pytestmypyESLintgitleaks
Growth & GTM
Marketing automationLead gen & segmentationSales funnelsEmail & SMSPaid socialConversion trackingGoHighLevelStripe

How I build

I work as the manager of a team of coding agents, not the person typing every line: I plan the work, the agents build against one shared policy, and I verify against tests and real behavior before anything ships. It's the same instinct I bring to a team: set a clear outcome, give people what they need to move, and check the result honestly. It's how I built the systems on this page, and it's a repo of its own.

How I lead

For 15 years I ran a company that was equal parts general contractor and healthcare provider, delivering some of the most complex accessibility projects in the country across residential and commercial sites. I kept patients, clinicians, insurers, case managers, attorneys, and the trades aligned on a single outcome, communicating clearly across technical and non-technical audiences and keeping every stakeholder's needs in view, the end-user's most of all. I negotiate hard, ambiguous situations to the finish, aim to be easy to work with, and leave every process simpler than I found it. That work earned a first-ever Heart & Hammer Award for service and patient care.

Projects

Selected work

A mix of portfolio builds and stripped-down versions of production apps I've shipped. Explore the live demos or dig into the source.

FeaturedBased on a production app
2026

Cairn

A privacy-first records system where an assistant can draft but only a human can commit - and every change joins a tamper-evident hash chain you can verify.

A sanitized reference implementation of a review-gated, audit-grade personal-records app. The AI can only propose a draft or a labeled assessment; a single private path commits a record, reachable only by an explicit human approval. Every action appends to a per-user SHA-256 hash chain - enforced in TypeScript by default and by Postgres triggers with row-level security - with a verifier that shares the writer's hash function. Zero-data-retention model calls through one gateway, versioned prompts, 50 offline tests.

  • Next.js 16
  • TypeScript (strict)
  • PostgreSQL + RLS
  • Responsible AI
  • Zod
  • Vitest
  • GitHub Actions
FeaturedBased on a production app
2026

RAG Report Platform

A production-grade engine for retrieval-grounded, multi-source AI report generation - provider-abstracted, layer-cached, and honest when a dependency fails.

A sanitized reference implementation distilled from a production app. Every model call goes through one offline-testable function; reports are cached by what actually determines them; and the model is never a single point of failure. Includes RAG over pgvector, an LLM-as-judge eval harness, idempotent billing webhooks, and row-level security enforced by a CI gate.

  • Next.js 15
  • TypeScript (strict)
  • PostgreSQL + pgvector
  • RAG
  • Zod
  • Vitest
  • GitHub Actions
FeaturedBased on a production app
2026

Audit-Grade ETL Pipeline

An audit-grade pipeline that turns inconsistent, multi-source financial documents into a reconciled, categorized ledger - and refuses to load anything that doesn't tie out to the penny.

A sanitized reference implementation of a forensic financial-document pipeline. One parser per institution behind a single interface, a penny-exact reconciliation gate that fails loud, SHA-256 idempotent ingest, a versioned categorizer that never overwrites human review, and redaction with a verification self-test. Integer-cents money end to end, mypy-strict, 51 offline tests.

  • Python
  • ETL
  • SQLite
  • mypy (strict)
  • pytest
  • GitHub Actions
2026

Agentic Dev Workflow

A portable, AI-native development workflow: one shared agent policy and a plan, build, and verify loop that travel across editors and machines, grounded in a reproducible Windows + WSL2 environment.

How I actually work - as the manager of a team of coding agents rather than the person typing every line. The portable core is editor-agnostic: a single AGENTS.md of standing orders read by Claude Code, Codex, and opencode, an OPINIONS.md of durable taste, and a plan/build/verify loop. This repo grounds it in a reproducible Windows + WSL2 setup with two bootstrap paths - imperative shell or declarative Nix (home-manager) - wiring up WezTerm, zsh, Starship, Neovim, and a multi-agent multiplexer.

  • Agentic Workflow
  • WSL2
  • Nix / home-manager
  • Neovim
  • zsh
  • Claude Code

Contact

Let's work together

I'm open to AI and full-stack engineering roles, remote-first. The fastest way to reach me is email, and I answer quickly.