Generate personalized cover letters from a job posting URL.

My problem

During my job search, I sometimes needed to write cover letters. Although I enjoy crafting the right phrasing, this exercise remains quite rigid and repetitive. That’s why I decided to make things easier for myself without undermining the process.

Automate…

• Delegate: the low-value part of the letter / the common part shared across multiple letters
• Save time: allow me to focus on revision and adding value where necessary
• Centralize: have a place to store all generated letters
• Test generative AI: I saw that Mistral offered free “agents” configuration on their platform, I wanted to see what it could do
• Have fun: take advantage of this small project to get started with frameworks/technologies I had never touched before

… with some constraints

• Easy to use: if the process is tedious, the whole principle is void - the goal is to make life easier
• No deployment: even though there are simple and free ways to make the service available online, my need doesn’t justify it
• Free: LLM usage could be expensive, so I need to find solutions that are (still) free and produce satisfactory results

Using the browser

I wanted something simple: the goal was to save time, so avoiding dozens of clicks or context switching. I naturally turned to a web interface: the links I would retrieve would be in another tab and I would just have to copy them into my interface.

For a while, I wanted to get into React - because I had the opportunity to work with Angular and Vue, but the project size made me lean towards Preact. It would be a good middle ground to get initiated into React’s philosophy. To bootstrap the project, I used Vite with Preact/Typescript/Tailwind.

motivation-mate

This is how motivation-mate was born, a web interface that generates cover letters from job postings. Just copy the job URL and the cover letter is created automatically!

Once the page content is retrieved, a request is sent to the pre-configured Mistral agent. This is where the heart of the application lies: I had to write a precise and personalized prompt so that this “agent” could produce quality content that was completely personalized. To do this, I had to provide information that I deemed relevant to highlight in a cover letter (my background, the technologies I’ve used, etc.). And finally, I provided input/output pairs (job posting content / cover letters specific to that content) as examples of what I consider a satisfactory result:

The response is then saved in a text file. You can then copy its content to use or modify it.

And there you go!

What’s next

But… That wasn’t enough for me. I wanted something even easier to use: this is how gogomate was born!