[technion]

The only advice any respectable cryptographer will have given is "don't invent your own crypto".

I'd urge you to reconsider having done so.

The problem with "please correct me if anything is wrong" is no one is going to invest significant man-hours reviewing yet another algorithm unless it comes with a detailed paper showing why it's better than the existing standard. Which it isn't, because there is no memory complexity here, and thus, Argon2 is a more interesting target.