DavidMendoza

The Problem

Bug reports followed a slow, reactive chain: someone reports, escalates to triage, then escalates to engineering. Engineers spent time on tickets that turned out to be false positives, and had to reproduce issues manually from scratch with little context. The whole loop was expensive.

The Solution

Automation that triggers Claude AI the moment a bug ticket is created. Claude reviews the ticket and flags it as a false positive or confirms it as real. For confirmed bugs: creates a draft PR, writes Playwright end-to-end tests and unit tests, records a screen video of the issue, and posts it to GitHub as a comment, so engineers can confirm visually before touching a line of code.

The Problem

Customers building event landing pages wanted to customize their layouts (change colors, resize elements, center content) but were not technical. Support agents in the customer service team were not trained in CSS either. Every question like "How do I change the button color?" became an engineering escalation, with days of wait time on both ends.

The Solution

Internal tool where support agents or customers paste their page URL and ask a plain-English customization question. Playwright scrapes the page structure to understand the layout, OpenAI analyzes it, and returns a specific CSS snippet they can copy-paste directly. No engineering required. Still in active use today.

The Problem

E-commerce sellers creating product mockups had a 5-step manual process per design: place the model, define the crop area, paste the design, export, then run through an image optimizer. That was 10 to 12 minutes per design. A seller with one shirt in 20 color variations had to repeat it 20 times. That is over 3 hours for a single product line.

The Solution

Electron desktop app for Mac and Windows. Upload a model once. Using Facebook's SAM2 AI, running locally with no token cost, define the product area in one click. From there, export unlimited designs in all color variations in a single batch. Also built: a WordPress plugin for automated licensing and activation, and a GitHub Actions CI/CD pipeline that packages and publishes new releases automatically.

The Problem

A non-profit organization had no way to accept online donations. Donor relationships were tracked in spreadsheets. Email communications to donors were inconsistent and written manually. End-of-month impact reports (which needed to go to donors and the board) took several days to compile and send.

The Solution

Built a full website with integrated donation platform, an email communication workflow for donor relationship management, an organization activity tracker that pre-populates end-of-month donor update drafts, and a one-click report generator for board-level communications.

The Problem

A growing English education business managed students entirely through email and spreadsheets. Enrolling each student was a manual, multi-step process. There was no visibility into student progress or attendance. Scaling the business meant hiring more administrative staff instead of improving the product.

The Solution

Built a full LMS with self-serve enrollment, integrated payments, student progress tracking, instructor dashboards, and automated milestone communications. The entire student lifecycle, from first payment to course completion, runs without admin intervention.

The Problem

In 2020, the pandemic shut down physical retail in Honduras overnight. Small business owners had no path to sell online. Existing platforms were inaccessible, too expensive, or did not support independent multi-vendor storefronts. The gap was real and immediate.

The Solution

Built, marketed, and operated a full multi-vendor e-commerce marketplace from scratch. Merchants could self-onboard, set up their own storefront, and start selling without calling anyone. Grew entirely through word of mouth, with zero paid advertising. Eventually wound down post-pandemic due to operating costs, but the model was fully validated.

The Problem

Existing personal finance apps show you data ("you overspent on food last month") but never explain the behavior behind it or tell you what to actually change. The insight stops at the chart. Generic budgeting advice does not account for how you specifically spend.

The Solution

Built a personal finance Android app, available on Google Play, that uses your own transaction history as context. An AI layer answers questions about your finances, identifies the behavioral patterns driving overspending, and gives personalized, actionable tips. Built entirely from a personal need.