His approach amounts to disabling safety features.

No, everyone is crazy and not thinking at all right now. Vector databases are a great example of cargo culting, as are many other approaches in AI and ML.

I increasingly work with the embedding vectors, but I keep them in memory or in a regular database column. By keeping them in a regular database you can tag ordinary records with locations within embedding spaces, and you gain all kinds of helpful clustering and joining capabilities through embeddings tuned to specific tasks. You just loop over the hydrated records. You get all the same benefits and more.

A working project counts for a lot. Kudos for getting as far as you have. As a person who has worked for startups for a lot of years now, I can tell you that your efforts count for a lot. I don’t know your background exactly, but I’m sure you have a lot to learn about a lot of things. Getting good at all of it can take the better part of a decade, and the senior people on your team know this. You’re likely safe in that regard. They’re not mentoring and helping because they’re busy, not because they disapprove or dislike what you’re doing. That’s just how things are, usually! In terms of making your project better, I personally rely a lot on logging to make the bugs and behavior more visible. DataDog is very helpful. Linting is another force multiplier that makes your code much less buggy and much cleaner for minimal effort on your part. Good luck, you got this!