Feed it websites you like, create a website you love
This personal project aims to explore and gain knowledge about the technology stack behind AI products. It also allows me to leverage my cross-functional expertise in product, marketing, and software engineering. Within this repository, you will find a well-documented journey that showcases my working style, process, and more.
The project will:
- Define a problem statement
- Validate the problem statement through market research
- Identify target audiences and a go-to-market strategy
- Define a technical architecture to support a public-facing proof-of-concept
- Implement a custom AI model
- Reflect on the final proof-of-concept
Check out my Miro Board here and my ChatGPT History here to see the journey!
I am excited to utilize the Storybrand framework, which I discovered during my time at Immutable while spearheading the go-to-market efforts for portfolio games. This framework empowers me to communicate a clear and impactful message, connect with users, and serves as a valuable reference when developing go-to-market strategies.
SKIPPING THIS for now - but a brief Google will tell you the pain points - website templates lack SEO, scalability, UX, customization. It's a boilerplate. Tentatively thinking that my target audience is people who use website templates, but also may expand to people who use Upwork etc.
Current scope -
- User puts in one to many website URLs
- User gets back ONE deployed and responsive website
The current goal is to test how bad/good we can train the AI model with minimal guidance. Since this is a learning experiment, it'd be good to have a simple goal, with very loose definition, and see how the overall results are.
Attached my current application structure. There needs to be some discovery -
- There seem to be existing AI models (ResNet, VGG) that will extract features from website screenshots.
- ChatGPT (the LLM) has been my main assist in project managing my approach to this project, and it suggests I need training and validation sets, but I don't know what that means.
- Lastly, my preliminary thoughts are that it would be useful to have the user train the model as well, not just the creator, but I'm wondering whether this would be a good user experience.
- This question should be kept in mind when defining inputs/outputs for the various services.
- This question should be kept in mind when defining inputs/outputs for the various services.
