Jorge Mora

HR teams were tracking candidates, roles, and pipelines in disconnected spreadsheets — no audit trail, no shared state, frequent mistakes.

Designed the data model and built the platform from scratch: candidate pipelines, role tracking, role-based access. One system that replaced the spreadsheet sprawl.

Hiring operations centralized before the team grew. The data model handled every edge case that came up in production without structural changes.

An optical retailer needed glasses detection from a live camera feed — with no cloud dependency and no latency from a server round-trip.

On-device CV pipeline: MediaPipe for face landmarks, a custom ONNX model at ~6.2 MB, WASM + WebGPU acceleration. Runs entirely in the browser.

Real-time inference in the browser. No video leaves the device. Works on a phone without an internet connection.

In Mexico, anyone can register a phone line under your CURP without your knowledge. There was no tool to check which lines were registered to you.

A web app that queries all major carriers and MVNOs in México with a single CURP lookup. Results in real time, no account needed.

A tool people actually share when they're worried about identity misuse. Still getting traffic.

NASA's public image archive is massive and the official search tool is too literal — it surfaces what you typed, not what you meant.

A cleaner interface over NASA's public API. Fast, filterable, designed to let you actually explore the archive.

Something worth bookmarking. The kind of project people share because they want others to find it.

A radio newsroom ran its entire operation across separate spreadsheets — shifts, payroll, commissions, expense tracking, loans. No shared state, constant sync errors.

One internal platform covering the full ops surface: journalist scheduling, payroll calculations, commissions, expenses, employee loans. Built for real workflows, not demos.

Spreadsheets gone. Three years in production without a major incident. Management got an audit trail for the first time.

A radio news station managed employee activities and social publishing manually — no unified view of what was happening, no automation.

A web app integrating WordPress and Facebook APIs: employee activity tracking, newsroom oversight, and one-click social publishing.

Daily Facebook publishing went from a manual multi-step process to a single click. Coordination overhead dropped noticeably in the first week.

Building something like Shazam from scratch to understand how audio fingerprinting actually works — CNNs, spectrogram matching, live mic input.

A real-time music detection system using convolutional neural networks trained on audio fingerprints, with a web frontend for live mic input.

It works. Not production-grade, but the core fingerprinting pipeline identifies songs from a live mic feed with reasonable accuracy.

Wanted to build something that runs ML inference in the browser from a live camera feed — ASL recognition was the right scope to make it non-trivial.

A real-time ASL alphabet classifier using a trained model served through the browser, no backend, just camera input and immediate output.

Runs entirely in the browser. Good enough to demonstrate real-time classification; good enough to build on.

Finding a free classroom or a professor at DICIS meant checking a printed schedule or asking around. Nobody had time for that.

Scraped the university's scheduling system and built a lightweight tool that shows available classrooms and professors in real time.

Students use it. Apparently it spreads by word of mouth — I find out it's still running when someone DMs me about it.

The thermodynamics lab at DICIS had no web presence — students couldn't find schedules, resources, or contact information online.

A clean web app for the lab: schedules, resources, staff info. Built with the engineering students as the actual users in mind.

The lab has a working web presence. Faculty and students use it regularly.

Producing daily weather forecast videos required manual editing — assembling assets, writing scripts, matching narration. Same work, every day.

A desktop app that takes weather data and automates the full video production pipeline: FFmpeg for rendering, Whisper for speech-to-text, NLP to structure the script.

Daily video production time dropped significantly. The output is consistent in a way that manual editing wasn't.

A small grocery store needed to weigh products and print barcode labels. Every solution on the market was either overkill or tied to a specific POS system.

A desktop app that connects to the store's scale, reads the weight, and prints formatted labels with barcodes. Nothing more, nothing less.

The store uses it every day. Replaced a manual process that was causing checkout errors.