AI deployment engineer, IT analyst, and co-founder. Eight years building production systems, the last three shipping AI across cloud, identity, and enterprise workflows.
Every system I have shipped, newest first. The one under your eye opens itself as you read down the log. No clicking, nothing hidden.
Bring your own keys. The platform never sees the plaintext.
A multi-provider AI workspace where the operator brings their own keys and the platform never sees the plaintext. Built and shipped as the sole engineer: provider abstraction across Claude, OpenAI and Gemini, retrieval on pgvector, OAuth for Microsoft 365 and Google, and AES-256-GCM encryption on every stored token.
Reads the policy library so the analysts do not have to.
An agentic identity and access platform that reads the policy library so the analysts do not have to. Document ingestion time cut 94 percent and total cost of ownership down more than 70 percent against the incumbent tooling. GuidePoint Security partnership in process.
A two-tier AI process audit, free in the browser, deeper when you pay.
A two-tier AI process audit tool: a free JavaScript tier that runs in the browser and a paid LLM tier billed through Stripe. Python on AWS Lambda with RDS and S3 behind it. Live at rjautonomous.com/ai-tools.
A Claude-native campus AI platform for Marymount University.
A Claude-native campus AI platform for Marymount University: retrieval over institutional knowledge with MCP-style tool orchestration on Azure. Phase 2 adds SSO-driven agentic ticketing so the platform can act on a request, not only answer it.
Built so that neither founder has standing access to patient data.
An electronic health record platform for therapy practices, built so that neither founder has standing access to patient data. PostgreSQL row-level security with a NOBYPASSRLS role, X12 837P claim generation, field-level AES-256-GCM encryption, TOTP MFA, append-only audit logging and encrypted peer-to-peer telehealth. Now in a live pilot with a funded design partner practice.
SimplePractice to QuickBooks, and PHI never reaches the model.
A billing automation pipeline for a therapy practice that was downloading CSVs, stripping patient data by hand and retyping financials into QuickBooks. Two-stream design: the PHI-free income report feeds QuickBooks journal entries directly, while transaction-level exports pass through a Python de-identification layer before Claude sees them. AWS RDS with KMS, CloudWatch and CloudTrail underneath.
Reads every inquiry, classifies it with Claude, routes it to the right person.
An IMAP monitor that reads every inquiry landing in the RJ Autonomous mailboxes, classifies it with Claude and routes it to the right person. Scheduled with cron on macOS, so it costs nothing to run and never sleeps.
Small tools for the work nobody sees. Built on the help desk floor, for the problems that eat an afternoon. Bought once, yours to keep.
Cisco VPN leaves files behind on macOS that convince the next installer a copy is still there. This finds every trace, shows you the list, and removes it.
Boots from a USB stick and erases hard drives and SSDs to NIST SP 800-88. Hardware secure erase first, verified overwrite when that is not available, and a signed record for every drive.
Turns one known good Windows deployment USB into up to five verified copies in a single run. Every destination checked on its own.
It is a real little shell. Browse the products with ls, cd and cat, or run scan and paste a job description to get a fit score read by Claude against my actual work.
What counts as work has been rewritten three times in human history.
When we left the fields for the factories.
When we left the factories for the offices.
And now, as we leave the offices for whatever comes next.
I believe we are witnessing the largest shift in human productivity since the Industrial Revolution. Not because machines are replacing people, but because intelligence itself is becoming a tool that can be designed, directed, and multiplied.
My work is simple:
Build systems that help people do more of what matters.
I do not know what late 2026 will look like.
I want to help build it.
Accepting partner inquiries. Tell me what your team does fifty times a week, and I will reply myself.
