Mori LiuProduct & Growth @ Hume AI | Creator (120k+) | prev. AI/HCI Research @ Columbia

ep. 2 of building a fashion app in public – backend edition

Name: ep. 2 of building a fashion app in public – backend edition
Uploaded: 2026-04-02T15:54:47+00:00
Duration: 1 min 21 s
Channel: Mori Liu
Description: In this solo walkthrough, the creator demonstrates how to add a durable, secure backend to a fashion try‑on browser extension by leveraging ChatGPT…

2 Apr 2026 · 1 min

Watch on YouTube

1:21

In this solo walkthrough, the creator demonstrates how to add a durable, secure backend to a fashion try‑on browser extension by leveraging ChatGPT throughout the process. The video starts by outlining the limitations of the current local‑storage‑only solution and enumerates the app’s user‑facing features. Those features are fed to ChatGPT, which proposes a relational database schema. The creator then validates each table for proper primary keys, appropriate data types, and security settings to prevent unauthorized access. Relationships between entities—such as a one‑to‑one link between a user and an avatar and a one‑to‑many link from an avatar to try‑on sessions—are mapped and expressed as foreign keys. The schema is visualized as a diagram for further verification. After a brief café change, the finalized schema is translated into SQL statements and a PostgreSQL instance is launched with in‑transit encryption enabled. Dummy data is generated for every table, and sample SQL queries are run to simulate real data flow, confirming that the backend functions as intended. The video concludes with the backend ready for upcoming frontend integration.

Chapters

Come with me to build a database for a fashion app using ChatGPT. If you're new here, I previously built a browser extension that lets you try on any outfit while shopping online. But although the browser's local storage persists between sessions and tabs, it lacks durability, which means if you clear your cache, get a new laptop, everything is gone. This is actually the biggest critique when it comes to bad coding because most people skip the backend entirely or build one that's not secure. So today, we're building a real backend while using ChatGPT to both map out the technical steps and speed up the process. Step 1 is to write out all the features for my app, aka what I'm allowing users to do. Then I asked ChatGPT to design the database capabilities based on my features. Once I got the initial design, it's really important to check whether each table has a primary key, which is its unique identifier, the correct data type for each element in the table, and finally, your level of security to prevent users from reading or writing data that's not theirs. The next step is to map out all the relationships between the tables, which determines how your data connects and talks to each other. For example, one user can have one avatar, which is a one-to-one relationship, and each avatar can have multiple try-on sessions, which is one-to-many. This will appear in the table as foreign keys. I asked ChatGPT to write out the schema for the database design and then visualize it as a diagram so I can validate further. However, the café I was sitting in was closing, so I went to another one that turns the to a bar at night. After I got my drink, I turned the final schema into SQL statements and spun up a Postgres database that automatically encrypts the data while it travels across the internet. The last step is to generate some dummy data for each table and run them through SQL commands to simulate real data flow. Once all the requests went through, I'm ready to call it a day, but stay tuned for the frontend integration.