Ghost for Beginners infrastructure
Public repoTurned Ghost installs into repeatable beginner-friendly deployment work: Docker-based local testing and AWS/Packer automation instead of hand-built servers.
Full-stack programmer and AI automation engineer. Currently leading automation and AI work across UC San Diego — instructional assistants, Oracle workflow automation, RAG interfaces, and screen-aware assistants — while shipping independent iOS and Cloudflare products on the side.
At UC San Diego I lead and contribute to automation and AI work across campus groups — reducing friction, improving workflows, and helping teams adopt AI in daily operations.
Recent threads: AI instructional assistant provisioning, RAG and natural-language interfaces, Oracle workflow automation, screen-aware desktop assistants, and agentic codebase research loops that evaluate and iterate on their own solution attempts.
Curated from GitHub, local repos, live app links, and internal-use systems. This leaves out scratch repos and keeps the projects that had real users, operational value, or shipped-product intent.
Turned Ghost installs into repeatable beginner-friendly deployment work: Docker-based local testing and AWS/Packer automation instead of hand-built servers.
Built draft trackers, data scrapers, and fantasy baseball utilities around real decision loops: rankings, team strengths, and fast draft-day feedback.
Made small Electron menu bar apps for NBA and Lakers schedules, then pushed the same pattern into Cloud9 and esports monitoring: focused utilities that sat where the user already worked.
Built a full scouting stack around demo downloads, match scraping, serverless APIs, backend jobs, and a frontend for reviewing Counter-Strike player and team signals.
Started turning internal codebases and team knowledge into queryable assistants, including 4D-focused retrieval work and smaller GPT-style tools for domain-specific lookup.
Shipped a golf swing analyzer with frame review, drawing tools, path tracking, slow motion, side-by-side compare, and a no-subscription positioning.
Built a run of focused local-first iOS products: reseller listing cleanup, wedding seating charts, prayer reminders, recipe workflows, and FODMAP support. The pattern was quick product wedges with real data models and release paths.
Created a local-first metrics system that pulls Google Analytics, Search Console, App Store Connect, and sales data into static dashboards and markdown reports for shipped apps.
Shipped private on-device photo enhancement for iPhone and iPad: upscaling, denoise, deblur, and a clear privacy promise with no photo uploads.
Shipped an on-device voice notes app with instant transcripts, smart summaries, search, no accounts, and no cloud processing.
Built and connected Google OAuth, Drive folders, Onyx/RAG setup, class document sets, and assistant provisioning flows for student and faculty support.
Continued moving messy operational workflows into working tools: Oracle billing/product ordering, UCSD accessibility checking, model-improvement loops, desktop assistants, and installer workflows.
RAG, LangGraph, LLM interfaces, agent loops, local speech-to-text, prompt-to-tool workflows, evaluation harnesses.
Production web apps, dashboards, reports, integrations, OAuth flows, and analytics — front to back.
Edge services, containerized workloads, cloud ops, and the boring-but-critical security & uptime work.
iOS launches, App Store workflows, pricing, analytics, launch copy, and privacy-first product decisions.
This page is the public résumé: links, shipped apps, and the kind of automation work I can take from a messy problem to a working system. Best route in: email.