Fullstack developer tests

These tests are made to assess your skills using the stack of our AI customer service app which:

1. Integrates with Gmail API, Shopify API, Airtable API, Webshipper API, OpenAI API
2. Generates automatic answers and actions to customer service emails, with human in the loop

Tech stack:

• Spring Boot & Kotlin
• Postgres
• Flyway for database migrations
• HTMX & Thymeleaf server side rendering
• Shoelace for UI components (web components)
• Gradle, Kotlin DSL (build.gradle.kts)

Other details:

• JdbcClient (new in Spring) for database access
• @Scheduled for scheduled jobs

Tests introduction

Tooling requirements

• Use Intellij ultimate. You can use it for free through: EAP program
• Use the built in DB client in Intellij to connect to the Postgres database (no external tools)
• Use the built in git client in Intellij to commit code
• Use the built in HTTP client for testing HTTP requests. It is better than Postman etc. It makes it easy to share version controlled requests in the repo
• Use the latest version of all libraries and tooling. AI might generate projects with older version, so make sure you fix this manually. This means the latest version version of Gradle, JVM etc. Many developers don’t know that you can run your Java/Kotlin code targeting an earlier JVM, but still run that bytecode on a newer (and more efficient) JVM. Show your project settings and show that you run on the latest JVM.

Make sure you use all these tools during the looms so I see you master them / know about them

Test - simple frontend + Backend integration

Make a loom showing and explaining all the code and the working web app in action. Use the same tech stack described above.

Make a spring boot app that has a page that:

1. Displays a button to load products
2. When the button is clicked, it fetches products from a Postgres database and displays them in a table.
3. Lets you add a new product via a form, which is then saved to the database and updates the table without a page reload.

The DB has been populated with a scheduled job that fetches products details from https://famme.no/products.json. Limit number of products you save to 50.

Note that one product can have multiple variants, so the json response from that endpoint does not necessarily translate cleanly to one database table. Solution: JSONB or multiple tables with foreign keys.

You also don’t need to save all the fields from the json response, only the ones you think is most relevant. Choose a few 3-5 fields at most to make your life easy.

Tips:

• Use flyway for database migrations (initial table setup).
• Use initialDelay=0, to make the scheduled job run immediately at startup.

Important:

• Make the table nice
• Use Shoelace for components like buttons
• Use HTMX to update the table without a full page reload

So the website would be something like:

1. header
2. button to load products into UI
3. table with products
4. Form to add new products (can maybe show up after table is loaded). Again, using HTMX to update the table without a full page reload.

Test - Use of AI

Tip: Make sure you use version control(git) when working with agents, then you can easily see what the agents changed

You need to show that you can use AI to create new features. I recommend using an agent like:

• Claude code
• Jetbrains’ Junie
• OpenAI Codex

So that the code base is edited directly in the IDE. But you can also copy paste code into chatGPT or Gemini and ask it to generate code. But that is more work and then you have to make sure you’re copying in the right context etc.

These are the features the AI should implement:

• A new page with a search box that lets you search for products by title. Active search that updates the table as you type.
• On the original page where you load the products, there should be a link to update the product details that takes you to a new page to update the product.
• Add a delete button to delete a product from the database, make sure to have a
  to confirm the deletion.
• One more feature, which you decide yourself

For every feature generated by the AI, explain the code generated and change it manually if needed if you think you can improve something.

The goal of this task if to see how well you work together with AI. And to show that you can see when an AI makes a mistake or where you can do better.