How We Built Catalyst AI: A Peek into Our Tech Stack and Journey to Signing Our First Agency

Hey IndyHackers!

I'm excited to share our experience building Catalyst AI, an AI-powered concept brainstorming tool for marketers and designers. Our main objective is to help creatives come up with better ideas and iterate faster.

We've just signed a contract with our first agency, and I thought this would be a great opportunity to share with you some of the technologies, methods, and thought processes we used in building our startup.

What is Catalyst AI anyway?
Catalyst AI works by using a fine-tuned base model specialized for ad creatives, that is used to build brand-specific models that create on-brand ad mockups.

Backend: Go
For our backend, we opted for Go, a powerful and efficient programming language that enabled us to get started quickly with minimal boilerplate and setup. The backend of the app is simple so there was no need to complicate it here.

Frontend: Vue and Vuetify
On the frontend, we went with Vue and Vuetify. Vue is a lightweight and flexible JavaScript framework that makes it easy to build reactive and maintainable user interfaces. Vuetify provides a comprehensive library of ready-to-use UI components built on top of Vue. We build out the UI using the Vuetify Figma component library which made is super easy to code. Instead of spending time on designing buttons we simply built it with the existing component, which was a matter of importing the right pieces of code.

API Integration: Stable Diffusion and OpenAI
To bring the power of AI to our users, we integrated with APIs like Stable Diffusion and OpenAI. We had previously used Replicate, but we found that these APIs offered us a better model selected and significantly better pricing.

Fine-tuning and deploying models
One of the game-changers for us was discovering Andromeda360, an incredible tool that helped us quickly set up the environment for fine-tuning and deploying our custom models (note: we're not affiliated with them, just big fans). This saved us from figuring out the whole MLOps thing.

Hosting and Domain Registration: DigitalOcean and Namecheap
Hosting and domain registration were other essential components of the strategy. We chose DigitalOcean for hosting, thanks to its developer-friendly platform and affordable pricing. For domain registration, we went with Namecheap, a reliable and affordable choice.

As we continue to grow and evolve, we're always on the lookout for new technologies and methods to improve Catalyst AI and provide even greater value to our users.

Since you made it all the way through the post I hope you don't mind a quick pitch: If you think Catalyst AI could be helpful for you or someone you know, please give it a try or refer others who might find it useful.

Want to give Catalyst a try? Sign up for the waitlist here: https://catalystai.design/ .

Cheers!