Gean Perez — Computer engineer

Portfolio demo · fictional brand · designed and built solo

Andrä Eyewear — B2B wholesale

Partners order and pay online; the warehouse picks from the same stock ledger—one product surface for sales, ops, and leadership.

Stack

RustPostgreSQLStripePayPal

Live demo

Explore live product

Storefront and wholesale consoles—the same build as the screenshots below.

Problem

Wholesale teams often juggle PDF catalogs, emailed POs, and separate stock sheets—so what was paid does not match what ships.

What I built

Public storefront, partner accounts, cart and payments (Stripe and PayPal), live inventory, warehouse queue, sales-rep console, and commercial admin dashboard—one Axum app.

How it works

Browse the public catalog
Partners sign in and submit wholesale orders
Payment confirms the sale and updates stock
Warehouse fulfills from the ops queue
Reps and leadership track accounts, revenue, and activity in their consoles

Outcome

One path from browse to paid order to pick list, with reps and leadership on the same data—no parallel spreadsheets.

Screenshots

Public storefront — Where buyers discover the catalogClick or tap the image to enlarge

Wholesale buyer portal — Logged-in partner building an orderClick or tap the image to enlarge

Sales rep console — Rep view of accounts and activityClick or tap the image to enlarge

Warehouse / ops — Queue of orders ready to pickClick or tap the image to enlarge

Commercial administration overview—metrics and controls in one placeClick or tap the image to enlarge

Partner login — Sign-in before the wholesale portalClick or tap the image to enlarge

Private product · end-to-end ownership

Trading Automation Platform

Strategies go live only after review. A read-only dashboard shows feed, process, and session health before and during the open—without turning the monitor into another trading terminal.

Stack

PythonLinuxpytestMakeGitHub Actions

Problem

At market open, several bots can run at once. A crashed runner, stale prices, or code that never passed review can lose money before anyone notices.

What I built

Supervised live runners, pre-open health checks, emergency stops, and a monitoring UI that cannot place trades.

How it works

Backtest and iterate on historical data
Promote only strategies that pass review
Start supervised runners connected to the broker
Orders execute under the approved rules
Track session health on the read-only dashboard

Outcome

Before the session starts, an operator can verify feeds, processes, and readiness. During the session, they watch health on a read-only surface—not a second order entry tool.

Supervised automation as a product discipline—not a claim about trading returns.

Screenshots

Operator dashboard — Session overview before and during the trading dayClick or tap the image to enlarge

Health check — Feed and runner status visible before the openClick or tap the image to enlarge

Session journal — Same session notes without opening multiple toolsClick or tap the image to enlarge

B.S. capstone · PUPR · 2022

ContentCat — Academic Capstone (2022)

2022 capstone at Polytechnic University of Puerto Rico (PUPR)—enter a handle, see scored tweets in the browser.

Stack

PythonFlaskTensorFlow

Problem

Faculty wanted proof the team could deliver ML as a product, not only a training notebook.

What I built

Web form, tweet fetch, model inference on the server, and faculty deliverables (poster, manual).

How it works

Reviewer enters a Twitter handle
Site fetches recent public tweets
Model scores each tweet
Results show in the browser

Outcome

A reviewer could enter a handle and see classifications in one sitting, with poster and manual as deliverables.

B.S. computer engineering capstone at Polytechnic University of Puerto Rico (PUPR), 2022—team deliverable shipped as an integrated demo for faculty review (screenshots from the finished capstone).

Screenshots

Capstone poster — problem statement and model approach (2022)Click or tap the image to enlarge

Result screen — Results view with per-tweet political leaning scoresClick or tap the image to enlarge

Home screen where a reviewer enters a Twitter handleClick or tap the image to enlarge

System overview — System diagram — fetch, inference, and browser UIClick or tap the image to enlarge

Gean C. Perez Gil

Andrä Eyewear — B2B wholesale

Problem

What I built

How it works

Outcome

Trading Automation Platform

Problem

What I built

How it works

Outcome

ContentCat — Academic Capstone (2022)

Problem

What I built

How it works

Outcome

About

Contact

Project write-ups

Problem

What I built

How it works

Outcome

Problem

What I built

How it works

Outcome

Problem

What I built

How it works

Outcome