Location: San Diego, CA
Remote: Ok
Willing to relocate: Yes
Email: raymond50romero@gmail.com
Resume: email me for my resume
Technologies: JavaScript/TypeScript, Python, React, Node.js, Express, PostgreSQL/MariaDB/MySQL, Sequelize, REST APIs, Mapbox GL, Vite, TanStack Query, Linux, nginx, pm2, DigitalOcean, AWS S3, GitHub Actions, JWT/bcrypt
Portfolio: <a href="https://gratmap.com" rel="nofollow">https://gratmap.com <a href="https://github.com/raymond50romero" rel="nofollow">https://github.com/raymond50romero <a href="https://linkedin.com/in/raymond50romero" rel="nofollow">https://linkedin.com/in/raymond50romero
Full-stack engineer, B.S. Math-CS from UC San Diego.
I was the founding engineer at Citmit, a construction-permitting SaaS startup. I joined when they had a product plan and no engineers, built the initial codebase and infrastructure from scratch, and led a team of four engineers through alpha. I integrated Stripe billing, an S3 document pipeline, and the application-side integration for their LLM feature.
Since then I've built and run GratMap (<a href="https://gratmap.com" rel="nofollow">https://gratmap.com), a live web app where service workers share and discover tip earnings on an interactive map. React + Mapbox frontend, Express/Node + MariaDB backend, on a Linux server I administer myself. I designed, built, secured, and deployed it, and I've been the only person on call for it since its July launch. Venue scores shrink toward the global median in proportion to how much data each venue has, then map to a tie-aware percentile, so a 2-report venue can't outrank a 200-report one. I ran the security audit before launch: breached-password screening via HIBP k-anonymity, rate limiting, JWT in httpOnly/sameSite cookies, ORM parameterization. I'm currently building an LLM feature into it.
This month I built CaseBoard with two teammates at an applied AI hackathon: a live dashboard for a personal-injury case, synced from Clio, with separate views for the law firm, the client's medical providers, and the client. My part was the Claude-powered assistant, the Python document search behind it, and the Next.js frontend with one teammate. Each person's data is assembled on the server for their role, so the assistant answers only from what they're allowed to see (a provider never gets settlement figures), cites the document pages it used, and when the file can't answer, drafts the question to the provider or client who can.
I work in Claude Code every day and can explain and debug what comes out of it.
Looking for: full-stack product engineer or founding engineer roles
raymond50romero · · focus · HN ↗
Full-stack engineer, B.S. Math-CS from UC San Diego.
I was the founding engineer at Citmit, a construction-permitting SaaS startup. I joined when they had a product plan and no engineers, built the initial codebase and infrastructure from scratch, and led a team of four engineers through alpha. I integrated Stripe billing, an S3 document pipeline, and the application-side integration for their LLM feature.
Since then I've built and run GratMap (<a href="https://gratmap.com" rel="nofollow">https://gratmap.com), a live web app where service workers share and discover tip earnings on an interactive map. React + Mapbox frontend, Express/Node + MariaDB backend, on a Linux server I administer myself. I designed, built, secured, and deployed it, and I've been the only person on call for it since its July launch. Venue scores shrink toward the global median in proportion to how much data each venue has, then map to a tie-aware percentile, so a 2-report venue can't outrank a 200-report one. I ran the security audit before launch: breached-password screening via HIBP k-anonymity, rate limiting, JWT in httpOnly/sameSite cookies, ORM parameterization. I'm currently building an LLM feature into it.
This month I built CaseBoard with two teammates at an applied AI hackathon: a live dashboard for a personal-injury case, synced from Clio, with separate views for the law firm, the client's medical providers, and the client. My part was the Claude-powered assistant, the Python document search behind it, and the Next.js frontend with one teammate. Each person's data is assembled on the server for their role, so the assistant answers only from what they're allowed to see (a provider never gets settlement figures), cites the document pages it used, and when the file can't answer, drafts the question to the provider or client who can.
I work in Claude Code every day and can explain and debug what comes out of it.
Looking for: full-stack product engineer or founding engineer roles