Justice C

Atinara LLC

Aug 2026 - Present • 1 mo

AI verification company. I built the whole stack myself: • Atinara.com: 27 pages in English and Spanish, deployed as a Cloudflare Worker, strict content security policy, no third party scripts, fonts or trackers, form spam rate limited. • A testing engine that attacks a client's chatbot with an in house built system with hundreds of prompts plus attacks a local model writes for that specific business: prompt injection, data leaks, made up facts, wrong language, drift, tone. Saves the evidence first, then writes the report. • Custom local judge and target models so client data never leaves the machine. • A crawler that sweeps whole cities by business type, finding 10s of thousands of businesses, finds which sites run a live chatbot or use AI, pulls contact emails, probes each bot for wrong answers, and drafts the outreach. • A test and monitoring layer with hundreds of automated test cases and health checks across the site, its forms, the engine and the model server, growing every milestone, refuses to report a number it has not measured.

Multilingual Web Developer

Jan 2024 - Present • 2 yrs 8 mos

Fast, secure small business websites in English and Spanish, delivered in days, not weeks: • Built and shipped a 27 page bilingual company site with one prompt in one day. Also have built over a dozen demo sites • Locked down by default: strict content security policy, no framing, self hosted fonts, no third party scripts, trackers or analytics of any kind. • Contact and testimonial forms with rate limiting and spam protection; submissions stored in Cloudflare KV and never passed to a hosted AI model. • A public verification registry page, brand marks in SVG, and print ready mockups. • 100s of daily automated health checks on the live site, and a same day fix auditing system

Independent Software Development

Apr 2023 - Present • 3 yrs 5 mos

Build and run my own systems end to end, working with AI agents under strict rules I write, using skills I've learned and developed from years of researching and learning web, software, and AI development • A private AI terminal in Python with a persistent event store, crash recovery, migrations and hundreds of automated tests that gate every milestone. No third party dependencies, no network by default. • A data system that pulls and stores hundreds of millions of records into a local database, running as a Windows service, with hundreds of automated tests and audits over the code and the data and a status page that only reports measured numbers. • Bulk data pipelines that archive hundreds of gigabytes, compressed and sharded, with statistical replays over more than a billion records at half a million records a second on a laptop. • Long running jobs built to survive anything: progress saved after every unit, safe to stop and restart, a status command that proves they are actually working. • Guardrails for AI agents: hooks that block risky commands, a lock that stops two heavy jobs fighting over the GPU, rules files that every session must follow. • Local model operations: custom judge and target models, 20 to 32 billion parameter models running on one machine, daily audits where a local model reads every file for defects. • Security by default: keys kept outside every repo, nothing sent to hosted models, no Google services anywhere in the stack. • Operating documentation, runbooks and plain language status pages for every system, written so a non-technical person can run them.