The value of a computer science internship: a Google Brain intern shares advice with fellow students
Nation-wide research conducted on behalf of the American Association of Colleges and Universities (AAC&U) indicates that 94% of the industry executives and hiring managers surveyed are more likely to hire a recent graduate who has held an internship with a company or organization than an applicant who has not had this experience.
Internships are especially important among computer science students who may not realize the ubiquity of computing across industries and the depth and breadth of opportunities that await them after graduation.
The Department of Computer Science at Virginia Tech is committed to providing quality experiential learning opportunities for all students, including internships both off and on campus. Of those students who plan to graduate from the department in 2022-23, 74% have completed at least one internship.
“We are always trying to diversify the pool of available internships to enable every student to attend at least one internship before graduation,” says assistant professor and CS experiential learning coordinator Mohammed Seyam.
Julia Costello, who serves as the strategic partnership coordinator and runs the CS|Source Career Fairs each semester, credits the hands-on aspect of internships with helping students apply the skills they’ve learned in class in real-time professional settings. “Internships add a key building block to academic learning for a successful career,” says Costello.
In fact, according to a recent survey of graduating seniors in computer science, 62% of those students who completed at least one internship were offered a permanent position at their internship location upon graduation.
The value of completing an internship is well-documented; however, the process of securing an internship might not be as clear. To help clarify the process, we spoke to graduate student Aditya Shah about his internship experiences as an intern with the Google Brain research team at their offices in Seattle.
CS@VT: Your research areas are Natural Language Understanding, Information Extraction and Multimodal Machine Learning, and Joint Language-Vision Model. Could you tell me more about them? How do they apply to your research at Google?
Shah: These are subfields under a broader umbrella of AI which uses text, images and text, or image and video and text, etc, and extracts meaningful information from them. At Google, I’m currently working on improving the underlying machine learning (ML) models used for Multimodal Document Extraction. For example, given a PDF document like a W2 or invoice, we would like to extract all the relevant important information from the document. Since we have different modalities like text or images of scanned documents, we call it multimodal.
My work [at Google] mainly focuses on improving the underlying machine learning models that Google uses for Document AI. I collaborate with DeepMind and other researchers to come up with a model that yields better performance for extracting information from the document.
CS@VT: What prompted you to apply for an internship with Google?
Shah: Google has always been one of the dominant players in AI research. The exposure to different projects, the scalability, and their impact are massive. You get to work on products that will be used by billions of people. Everyone here is extremely talented. All of this makes Google an ambitious choice, especially for AI research.
CS@VT: What advice do you have for students who want to pursue similar research internships?
Shah: This can be quite subjective depending on the domain. For machine learning and artificial intelligence, having strong publications and relevant research experience can make you stand out from other applicants. I won’t say it’s impossible to get a research internship without a good publication, but it’s very, very competitive, especially in big tech companies. So you can start by finding out your area of interest. In research, it’s important to identify that one area of niche. Reach out to different professors for a potential research project and try to publish good papers. Along with this, make sure you have good knowledge of current research trends going on in your field.
CS@VT: What are your main takeaways from Google Brain’s application process?
Shah: Getting an interview call for an ML research role as a “non-Ph.D” student can be more challenging than actually cracking the interview. So, if your profile aligns well with the role, reach out to people for a referral. A referral from a relevant person can go a long way in getting a call for competitive roles. For ML research roles, sound knowledge of ML and math with relevant publications is a plus.
In addition to our interview, Shah answered student questions about his internship experience, which can be viewed on the department's Youtube channel.
This interview has been edited for length and clarity.
Written by Tayler Butters, CS Communications Intern.
Shah's Application Experience
With a referral from a Senior Research Scientist at Google Brain Shah was invited to a total of six interviews:
- Three interviews were general in nature. They were open-ended and primarily focused on his past research and current trends in natural language processing, deep learning, research design, and math.
- Afterward, he completed a project preference questionnaire and was matched with a project and project manager.
- The final three interviews were specific to the projects he selected. These interviews were based on project details and were designed to gauge his familiarity with the research involved.