Recently, in my spare time, I've completed a small platform for developing and deploying SaaS solutions.
What I have done is basically I wrote and deployed stuff that is required in pretty much any SaaS.
- Authentication: both email/password and Google login. Tech: cognito.
- Database: where user info is stored (e.g. subscriptions). Tech: dynamodb.
- Frontend: standard pages and flows: landing, pricing, account and workflows like login/register, change password, Google login, delete user + translation into several languages (besides English). It is connected to backend. Tech: vue3 js.
- Backend: with auth and stripe integration and some other stuff (e.g. sending emails, domain integration in prod account, counting jobs done to bill the customer etc.). Tech: aws python lambda (with aws powertools) and api gateway.
- Payments: stripe with three plans, all already working with frontend and backend, with functionality to change/cancel etc. (I use stripe customer portal). Tech: stripe.
- CI/CD: with email notification on failure. Push to dev -> rollout to dev account, main -> prod account. Tech: cdk pipeline and sns.
- Some E2E testing. Tech: cypress
Additionally:
- there are 3 AWS accounts with SSO (main, dev, prod). Tech: AWS Organizations/AWS Identity Center.
- IaC. I tried to minimise the manual steps as much as possible. Tech: CDK in python.
- Billing alert: when AWS charges exceed $10 per month.
- Everything is serverless and I pay almost nothing for hosting and developer use (because there is no traffic). I pay $2-5 per month now, and that is mostly for CI/CD (AWS Codepipeline, CodeBuild) as I am actively developing.
What is NOT there is THE PRODUCT ITSELF.
I started researching some problems I could potentially solve with a SaaS, but I am not very good at it.
Eventually I compiled a list of directions:
- LLM with OpenAI api calls (and vector database): since everyone is doing it, I thought this would at least be a good experience.
- Platform: I've posted similar questions before (about 9 months ago) and some people suggested to sell a platform of what I've built. The thing is, it is a code-first platform, and you need to be a developer to extend the platform for your needs. And as far as I know, no-code/low-code tools are what sell.
- Some other stuff: things like document parsing, employee time management, post-quantum encryption.
I could not come up with anything better for now.
Question:
Any ideas on how to go about it? I've read books like "moms test" and "zero to sold", but I'm not quite sure how to start correctly here. As far as I understand the most important thing is to find a painful problem, which I have no idea on how to do.
Current idea:
My current idea is to start with LLM OpenAI MVP, if that fails then go for the platform (I will also use the experience I got from the LLM project to make the platform better). And if that fails too, I will create educational courses on those.
PS:
I am based in Germany, that's why I translated frontend to several languages (i18n library).
Thank you for your input
Solve a problem instead of applying technologies.
Or just enjoy the process of creation for being the process of creation and accept that you made a thing that doesn't have actual utility to other people just for the sake of making the thing that should be made.
I mean you could hang it on the wall, call it art and move on to making something else. Or not. Good luck.