I think machine learning is not taught enough relative to it's growing demand.
COBOL is a language that is still used in many places and will need people to replace those who are retiring, so I feel that is lesser known and not taught in many places.
- Automata theory
- Complexity theory (P vs NP stuff, not the one from leetcode problems)
- Verification
- Optimization using stuff like genetic algorithms, ant-colony algorithms, particle swarm, simulated annealing, etc. Not sure how to call that area. In my case it was an AI course.
- Fuzzy logic
- Scientific computing and Numerical analysis (CFD, finite elements, etc)
- Information theory
Yes please, if we could have more software engineers working on mechanical/aeronautical/civil/<physical> engineering tools that would be great.
Currently the state of the art takes a long time to get out of PHDs and into industry, and when it does it is in the hands of a few companies that sell expensive software licenses.
The likes of at Ansys, Siemens (NX), CD-adapco,
You can access FOSS tools like open-foam but the average engineer doesn't have the skillset to also wrangle the tool as well as their domain problem.
An area I was always fascinated with but never had chance to dig into it was topology optimization to generate structures (or flow paths) to accomodate some given boundaries and physics.
That was my favorite subject in college and I wanted to pursue that area but sadly we didn't have any experts in the CS department. All of them were from the Mechanical Engineering department and they were extremely disconnected.