I am a RTL frontend designer by training. By my pure passion (and luck) for AI since 2015 I was able to get into AI HW accelerator space. I worked for 2 years making FPGA accelerators after which with the help of Udacity I moved within the company into building deep learning models. I worked on NLP models and also doing workload analysis for emerging hardware. But have to leave this team because old manager was a jerk. (She left/fired 8 months after I left). Now I am in architecture team for emerging hardware and working on building hardware simulators.
Here are few questions I have
1) I have a broad range of skill set from building deep learning models, building architecture simulators, mapping algorithms to HW and writing RTL for HW. What is the best way to use this skillset? I often have the big picture in the team but my team/org doesn't want to use this expertise. What should I be doing ?
2) I am very passionate about doing research in deep learning especially at the intersection of NLP and vision to build artificial-agents which can reason using language as latent-representation. I find any of the research jobs pretty hard to crack without a PhD. All my work done in the old team never made anywhere because of jerk manager. I am in process of building a github profile but main question is
Should I go get a PhD to work on these research problems? or I can get into research teams in industry?
3) From 1 and 2, should I pursue what I love viz doing research or suck it up and use my breadth of skillset to get into leadership positions?
4) This question is for H1B HW engineers who became SW engineers.
a) Was the H1B transfer straight forward without questions on HW skillset vs SW skillset?
b) How did you manage the skillset verification for PERM and I-140?
YOE: 6 Years
- Adobe F📴moreCan confirm about Stanford AI courses - the SCPD ones are quite good and dont require any criteria. If you are more serious in doing a Ms, would recommend OMSCS
- Given where you are, I'd not recommend a PhD at this point (I have one). The cost of lost income vs return is not enough imho.
Why are you thinking about research teams vs product development?
- If you wanna do research, you could potentially go to a university. Do a PhD and then teach and do pure research.
It does not pay as well though.
In the corporate world, there are a few places where research without having to worry about deployment works, but I dont believe that will lead to any significant career growth.Sep 16 1
- Following this, as I’m in HW myself looking to move to AI SW positions in future. I would say, go for PhD if you are really passionate about the subject, and not just to get a research job in one of the companies. Easier thing to do will be taking relevant part-time courses, leetcode and keep applying.