Data Analytics Engineering Intern
Department: Product Development Engineering / Data Analysis
My experience
Overview
Created a tool to identify and group failure shading patterns on semiconductor wafers using unsupervised machine learning. Applied self-supervised/contrastive computer vision to large, unlabeled data wafer data. Lots of data wrangling using Python. SQL + Pandas for tabular data queries. OpenCV and PIL for image processing. Machine learning was done using PyTorch, specifically PyTorch Lightning. Tools used most often were Anaconda, VS Code, and Git. About half the time was spent doing the dirty "data cleaning" (fetching data, inspecting + cleaning), and half was spent modeling (making and analyzing machine learning models). Towards the end, I had to make a front-end for the tool, for which I used Streamlit.
Having a good manager and team is just as important as enjoying what you do. I was very, very fortunate to not only enjoy my work, but also enjoy who I was with. Consider who you'll be working with before you accept a full-time offer. Good money doesn't mean you'll love your job!
Pros
Very open-ended internship. Lots of creative freedom on my particular project. Work was very challenging, interesting, and fulfilling. My team was amazing -- very knowledgeable and personable.
Cons
I had to be the expert; no one could really help me out on the machine learning side of things since I wasn't on a software team. Company was slightly behind in its AI/machine learning culture compared to other semiconductor companies like Micron or Intel, but the point of bringing interns in is to bring in fresh people with new ideas.
Impact of work
Time spent working
How did working remote affect your experience?
Hybrid work schedule. Most full-time engineers in my area came on-site Mondays and Wednesdays, then worked from home the rest of the week. Interns were free to work on-site or remotely as appropriate, which was nice. Sometimes, it felt empty when the rest of the team was remote and you were the only one from the area on-site, but it was still easy to communicate with everyone.
Interview advice
How did you find the job / apply?
Interview Rounds
Interview type
Interview questions
Lots of behavioral probing. I remember my manager asking, "Who are you, really? What gets you up in the morning?" And he wasn't looking for a corny, scripted answer. He wanted genuine answers about what you valued, where you wanted to end up in a few years, and why you were interested in the company. There's no way he would've hired anyone if they just came for the data science but didn't care about the product (semiconductors). I personally liked this style of behavioral interview, but it isn't the case with all hiring managers at this company. I actually had an interview with another team (same company, same place, just for a Process Engineering Internship) which felt very... scripted. They'd go through the 5 company values and ask you to talk about a time in your life you exemplified that value (i.e. "Tell me about a time you were innovative/trustworthy/inclusive"). Other teams had multiple rounds of interviews. I just had a single interview with an engineer plus an engineering manager, both of whom I worked with throughout the summer after I was hired on. Lots of questions about past experience to get a feel for what I was capable of. They specifically asked about times I used supervised and unsupervised machine learning in the past, so I talked about projects I did. They went through my resume and asked me what I did for basically everything that was listed on there, and they were looking for how I made an impact. Talking about how you made an impact was the hardest part for me personally. Theoretical/technical screening was way easier here than most other data science internships would probably be. There was a pretty simple theoretical screening that ramped up in difficulty. They were all conceptual, so you didn't have to list out equations or anything; just give a general definition and say what the thing is used for. The questions I remember were to define the coefficient of determination, define a P-value), talk a bit about differences between supervised and unsupervised machine learning, and finally to talk a bit about bootstrapping and random forests. My manager actually misspoke and asked me to define what a "stacked forest" was instead of a "random forest," which kinda flustered me. I had a really long awkward pause, but I just tried to do my best. Technical screening was very, very simple. No coding session, just a verbal pseudo-code session. Questions were how you'd read in a CSV with Python using Pandas, how you'd go about making a pareto chart for only certain columns in tabular data using Python. They were just using these to gauge your proficiency in basic Python packages like Pandas and Matplotlib/whatever plotting library.
Advice on how to prepare
Basic technical qualifications were just fluency in the data science side of Python, specifically with some experience in machine learning. Take a course or two in those areas to have projects to talk about; in my case, I also had prior work experience that I talked about during the interview. Although hard skills are obviously important, my hiring manager straight up told me on my last day that the behavioral fit was more important. Be yourself during the interview. Practice or go over some interviewing tips with career counselors at your university or elsewhere.
More questions? Send the reviewer a message!