Ricardo Reyes
Ricardo Reyes

Ricardo Reyes

Product development & data analytics

San Francisco · USF, December 2026

I'm the Technical Lead & E-commerce Manager at Giraux Fine Jewelry in San Francisco. I work with the owners on operating priorities and build the systems behind our storefront, bookings and internal tools.

Outside work, I build with SQL, Python and AI development tools. My projects cover transit data, financial planning and tools for other developers. I also love working out and video games.

I'm finishing my Business Analytics degree at USF in December 2026.

Available January 2027 · Open to US relocation

See my workResumeEmail meGitHub

Work experience

Giraux Fine Jewelry

Apr 2025 to present

Technical Lead & E-commerce Manager

  • Proposed and built our Cloudflare media platform ahead of a vendor upgrade. Recent Cloudflare invoices averaged 93% below the planned $89/month Sirv price. That comparison covers four invoices issued July to October 2026; maintenance and development time are separate.
  • Built the SimplyBook.me booking integration. Monthly bookings grew 59% year over year, with 1,200+ appointments recorded.
  • Built a Python pipeline to convert NetSuite exports into Shopify imports for 849 products and 859 variants, including validation of 9,489 image URLs.
  • Work with the owners on priorities and improve the tools after staff feedback. Documented 111 shipped items across 13 weeks of owner updates.
Read the media platform case study
giraux.com homepage
giraux.com

Selected projects

Giraux media platform

Giraux Fine Jewelry, in production

Replaced a media vendor with a platform our staff use daily. Migrated 476,851 files to Cloudflare Workers and R2, cut spin-viewer page weight 97%, and reworked bulk editing after staff feedback.

TypeScript · Cloudflare Workers · R2

A diamond in the 360° spin viewer
drag to spin

Technology cost and R&D forecasting

Independent portfolio analysis

A 12-month Giraux cost model with an SQL invoice ledger, maintenance assumptions and a keep-or-switch decision. Alongside it, an Uber R&D forecast uses six quarters of public filings to compare spending scenarios.

Explore the models and source

Python · SQL · Google Sheets

gtm-warehouse

Portfolio project

Built a SQL/dbt warehouse on DuckDB for outreach funnel and reply-rate reporting. Implemented SCD Type 2 lead history, incremental reply processing with a 14-day lookback, and a daily schedule. Validated the build with 76 data tests and matching row counts and content hashes across consecutive runs.

Run the synthetic example

SQL · dbt-core · DuckDB · Python

SF Muni transit equity analysis

USF BUS 410 project

Built a Python/DuckDB pipeline joining GTFS transit data, Census income data, and tract boundaries to analyze 28.7 million delay observations across 68 Muni routes. Built the React/Mapbox website and data visualizations, including a live Muni tracker.

Live site

Python · SQL · DuckDB · React · Mapbox

Map of San Francisco with census tracts shaded by median household income and 68 Muni routes colored by on-time performance

aidep

Open source

Built a tool that scans repositories for deprecated AI APIs, opens migration pull requests, and checks the changes against the repository's prompts.

TypeScript · GitHub App · PostgreSQL

Signet

Solo product

AI compliance proxy: PII scanning, HMAC-SHA256 tokenization, and generated GDPR Article 30 processing records. Next.js marketing site with a free compliance self-assessment.

TypeScript · Hono · Fly.io · Next.js

One proxied request: the raw customer fields on the left, the tokenized request Signet sends to OpenAI on the right

SF Accountability Ledger

Built a Python pipeline to enumerate 58,905 sf.gov CMS documents and assess publication status and discoverability for 12 reporting obligations, with automated tests and change monitoring.

Live demo

Python · Public data

SF Accountability Ledger with columns for reporting obligations, responsible departments, due dates, filing status and findability

Hiring-bias analysis with XGBoost

USF team project

Evaluated an XGBoost hiring classifier in a four-person course project. Found 70% accuracy below a 71.5% majority-class baseline and examined referral and employment-gap disparities.

XGBoost · scikit-learn · pandas