GRASP Lab 2024 Summer Internship Round-Up

Zoox, in Foster City, California

I interned at Walt Disney Imagineering as a Ride Control Engineer at their main campus in Glendale, California. I relocated to the Los Angeles area for the summer.

Microsoft (AI Frontiers team), located in New York, had to relocate.

I interned at ETH Zurich in the Robotic Systems Lab as a Robotics Student Fellow. The internship took place in Switzerland, and I was provided with a stipend to support my relocation.

NVIDIA, Onsite in Santa Clara, CA

SpaceX at Hawthorne, CA. Yes, I had to relocate to the Greater Los Angeles area for the summer.

Cruise LLC (self-driving car company), the internship was remote, no relocation was needed.

First on the internet, but 2 months after I submitted the application, their university relation team came to Penn, so I asked to speed up my resume being reviewed. I got 2 rounds of interviews, one for generic programming, one specifically for machine learning coding. After those, I talked with the hiring manager and HR respectively. Since salary and compensation were not negotiable, I just accepted the offer.

I first heard about the internship through searching online for engineering companies where the main product was not specifically an engineering focused product or service. Disney theme parks provide experiences that are unlike any other and can bring magical moments to people of so many different backgrounds and this involves some engineering. The application and onboarding process was straightforward and I felt like I was well guided to understand how to start working with many creative people who are not in engineering for the delivery of specific tasks.

A researcher from Microsoft gave a talk at UPenn, and I chatted with them afterwards. They passed my profile along to their team, and I was later contacted for an interview.

I first became aware of the fellowship through an email circulated by the GRASP lab during the first semester of last year. The application process involved submitting a statement of purpose alongside my academic transcript. Applicants were also asked to list their top three preferred labs, and selections were made based on the labs that expressed interest.

Found out about it on LinkedIn. Reached out to the hiring manager for the post, and he scheduled a couple of interviews. One of them was to test my knowledge of deep learning concepts and the other one was talking about my research.

I heard about and applied to the internship opportunity through the company’s website. It was interesting that they only had a single internship application for the all engineering internship positions (except for software engineer positions), and they matched me to a team based on the skills and experiences in my resume.

They started their recruiting process in April 2024. I had two rounds of interviews with the first one being a coding round and the second round was presentation-style where I gave a talk on my research. The onboarding was smooth. I received an offer on May 15th after the interviews were completed and got the opportunity to work with a mentor who actually interviewed me. The internship was from June 10 – August 30.

Applied ~50 positions, 3 of them went into the interview process. Only Zoox gave me an offer 🥹

I specifically chose this internship because of the unique opportunity for engineering to contribute toward a creative project and goal. The work at Walt Disney Imagineering is heavily driven by the creative process of developing new experiences to inspire emotions within guests and for me that was a very exciting application of robotics and automation to assist with creating the magic to leave guests with a lasting memory.

I applied to around 10 companies. I chose Microsoft because of the unique opportunity to work on an interesting problem with great researchers and large computational resources.

This was the only fellowship I applied for, as I was particularly eager to gain research experience outside the U.S., aligning with my long-term goal of pursuing a PhD. I chose the Robotic Systems Lab specifically due to my research interests in legged robotics and autonomous navigation in complex environments.

None. I chose this internship since NVIDIA was working on a lot of cutting-edge research that I was interested in, and the team culture seemed to be a good fit from my interaction during the interviews.

I applied to around 50+ internship opportunities. I chose this internship at SpaceX because it seemed to be the most relevant to my robotics and control background, and because I was able to ask about the potential projects I would be able to work on during my interview. The hardware I got to work on seemed to be similar to some parts of my thesis.

I applied to many other ones and interviewed for a few of them. But none of the other companies were hiring for a project related to my area of research. That was my most important criteria. At Cruise, I worked on a project that used machine learning as a tool for trajectory evaluation from self-driving cars. This project was in line with my research.

Yes, I did. As long as you get the offer, you could start the process as early as possible.

Yes. I went through the CPT process for my internship. The CPT process is relatively easy to navigate; the CPT webpage and Tori‘s emails are more than enough information to have a smooth onboarding into the process. My only advice would be to secure the CPT letterhead from the company well in advance before you begin the CPT process.

Yes, I’m an international student. I applied for CPT through UPenn and the process was very quick. It is very easy to navigate. As a PhD student, I had to first get a letter of permission from my advisor. After that, email the necessary information such as offer letter etc., to your department coordinator. From there, it is pretty smooth. The advice I would like to give is start early as soon as you receive the offer letter.

Reach out to HR teams if possible!! It should be more effective than finding someone’s referral than just submitting your resume.

I would suggest particularly for Masters students that they identify what similarities and differences there are between the day to day work of research and industry opportunities to then see what skills would be best to focus on.

Networking is important. Attend talks, conferences, and chat with industry researchers you find interesting. Be able to talk about your research and your contributions in a crisp and clear manner.

As some general advice: it can be daunting to apply for the same internships as your peers, especially in such a competitive and high-achieving environment. However, I think it’s really important not to let fear of competition hold you back. It can be very discouraging to receive rejections but all you need is one acceptance. As my mother used to tell me: “You already have a ‘no’ if you don’t try—your only chance at a ‘yes’ is by putting yourself forward.”

In my opinion, the key to securing internships is to expand your network. This network can typically be done in conferences, symposiums, or meetups. I’d also recommend being active on platforms like LinkedIn, or mailing lists, since recruiters sometimes look on these websites and sometimes reach out to you directly too (with internship/full-time opportunities)!

Talking to more like-minded people about your research interests also helps in finding relevant leads on industry/academic teams that are hiring people with a particular skillset.

For the summer of 2023, I applied to only a few internship positions (less than 10) trying to find the right ones that seemed most relevant to my research interests, and I wasn’t able to get any offers. I recommend expanding the pool when looking for potential opportunities, especially if you have to apply without any referrals or connections. I never really knew of the team nor the department I worked at this summer, but they were specifically looking for an intern with a robotics and control background with Python and LabVIEW experience, and I was fortunate to be able to matched to the team based on my skills and experiences. Being able to demonstrate your relevant skills clearly in your resume with keywords seem to be crucial because your resume is the one piece of paper that they use to determine your capabilities.

It was hard in my experience to find a research internship in my area of interest since there’s a lot of shifts today in the industry towards LLMs. However, if you search hard and persevere, you will find one. It might not be at a company that you expected, so keep an open mind and keep applying. Referrals are definitely useful. For my internship at Cruise, I was referred internally by a Penn alum whom I knew well. Use your connections to have a more productive internship search.

Machine Learning Engineer. Specifically our team are building an RL planner for autonomous robotaxi.

My daily responsibilities involved working on the system and hardware design for a new and upcoming attraction. This primarily involved control work dealing with sensors, PLCs, and fundamental system analysis. There was a particular emphasis on efficiency, repeatability, and safety due to the required uptime for daily operation with guests.

Coming up with research ideas on how to better train Large Language Models, and then implementing the ideas. We were heavily inspired by deep reinforcement learning and planning.

I worked on human preference alignment of text-to-image models during my internship. This would primary classify as computer vision. My daily responsibilities evolved during the course of my internship. The first couple of weeks was spent in doing a thorough literature review, and implementing extensions of a baseline. This was followed by systematically studying the limitations of the baselines, which were plentiful owing to the relative nascency of the field. I formulated and implemented some improvements, spent bunch of time performing ablations, and finally wrote a technical report to conclude the study. I had some follow-up ideas, but we ran out of time and I implemented a proof-of-concept before wrapping up my internship.

Mostly Programming (Python & LabVIEW) and a few hands-on experiments.

This summer, I worked as an AI PhD Robotics intern at Cruise Inc. within the Behaviors Organization. My primary responsibilities involved grading the road performance of autonomous vehicles (AVs) post-operation using machine learning (ML) models. These models are crucial for evaluating whether changes to the AV’s policy—specifically the end-to-end decision-making pipeline beyond the perception and prediction layers—result in performance improvements or regressions.
During my internship, I identified two significant challenges in the current setup:

1. The difficulty in validating outputs from the ML model for data without labels.
2. The lack of interpretability of labels derived from human label preferences.

To address these challenges, I proposed a novel problem formulation that reduces the dependency on human label preferences, which often lack explanations for their choices and don’t correlate with high-quality metrics. These metrics are essential for scoring nuanced interactions between the AV and other agents in the scene.

9am – 6pm. The working hour is flexible, and the reason I stayed so long is that Zoox offered breakfast, lunch, and dinner.

I worked 40 hours a week and outside of regular hours there were some opportunities for site visits and tours to better understand how Disney parks operate.

I liked to come into office around 3 days a week, for in person meetings and talks. The internship was 3 months.

I did not have set hours. I was expected to work 8 hours a day on weekdays. There was no pressure at all to work on weekends, contrary to the typical PhD experience. There was no pressure to work overtime as well, which gave me a lot of time outside work to explore hobbies and spend quality time and having a healthy work-life balance.

SpaceX had a 12-week minimum for all interns. I didn’t have set hours, but I usually worked from 9:30am-7pm to be able to have enough time to work on my projects.

I had to work up to 40 hours a week, but the workload was pretty light. I could set my own work schedule. Since the office was located in San Francisco, their usual hours were in PST. However, I worked in EST.

Yes, at GRASP I have had the ability to explore my research interests at the intersection of embedded systems and control in the Kumar Lab and this engineering design process and collaboration helped prepare me for working with experts in backgrounds different from my own. Being able to get excited about someone else’s work despite it not being my primary area of focus was especially helpful for having conversations with many of the people I have worked with at GRASP and at Walt Disney Imagineering.

Yes, definitely. I learned most of my machine learning skills from doing research at GRASP.

It definitely did. During my time here so far, the coursework and research has helped me build a very strong foundation and expertise in deep learning, computer vision, and optimization. One important skill I’ve mastered over my time here is “learning to learn fast” which helps me pick up any new topic or programming language/library very quickly. Working at NVIDIA, I had to use their software stack and work on a relatively new problem (text-to-image alignment) which required picking up the skills and catching up with literature as quickly as possible, which significantly reduces the onboarding time, so you get to spend more time doing actual work.

Yes, the first time I learned how to use Python properly was through MEAM 6200, and one of the main technical expertise I was complemented on was Python proficiency. Other soft skills such as the collaborative research experience I had at GRASP helped me in working with technicians and engineers in other departments within the company.

Absolutely! I think my mentor hired me because of my experience in using machine learning for one of my research projects. The skills I learned were good software engineering practices which now helps me write better code during the PhD.

There are lots of social events for interns. We could know what other interns/FTE do during seminars, free food hours, and deepen our relations in activities like top-golf, farewell cruise (YES!!! a cruise event for interns!!!)

Working to understand how guests with disabilities interact with rides in themed entertainment was particularly interesting for me. It allowed me to view themed entertainment attractions through a new perspective to see that improved safety and efficiency doesn’t just satisfy requirements but also enables a more inclusive experience. This had always made sense to me in the final product, but it was new to me in the design phase of the project which pushed me to be even more creative with my technical focus.

Being able to learn from the world class researchers at Microsoft. Living in Manhattan was also very fun.

The work culture was a refreshing experience and nice change of pace from my PhD, and working on a completely fresh problem gave me lot of new insights. I like to learn new things, and this internship was very intellectually stimulating. The team was very supportive of my work as well which was very gratifying.

My favorite part of the internship was that I was able to learn a lot thanks to the various tours and Q&A sessions SpaceX offered for their interns. I was able to learn so much about the commercial aerospace industry in general, the intricate design of the Raptor engine, control strategies of reusable rockets, and various nondestructive evaluation methods for inspecting aerospace vehicles.

The exploratory nature of the project. I loved not being told exactly what to do but rather given a current version of a solution and find innovative ways to improve it.

Yes, I learned how to implement and train Large Language Models, and I think this is a valuable industry skill.

For a lot of people, writing “research-grade” code blunts coding skills in terms of real-world usability. I worked on the NVIDIA NeMo open source library, meaning I had to adhere to style guides, readability, doing code review with peers, in addition to the research work. This I believe is a very important skill that I got to brush up on during my internship. The internship also prepared me for a potential future career in industry, by providing me with a better perspective on the life and expectations from an industry career.

Yes, one of the key skills I gained was time management and communication during collaboration. Since I had a lot of programming tasks where I had to build on top of existing projects or APIs, I learned how to effectively plan out my intermediate milestones for my collaborators to be able to contribute their portions within a limited timeframe. I also worked on and tried to improve on keeping project stakeholders up-to-date to make sure there aren’t any misunderstandings or confusions along the way. I have already started incorporating these methods in my collaborative projects at Penn, and it seems to be very helpful in making sure everyone is aware of my progress. The internship definitely helped prepare me for a career in industry because I was able to learn more about what it takes to be a responsible engineer for a particular system, and it gave me the confidence that I am, in fact, capable of making the right engineering decisions.

The skills I learned were good software engineering practices which now help me write better code during the PhD. I definitely think that this is useful for both industry and academia.

Yes, I would be interested in returning to the themed entertainment industry because I really enjoyed the engineering work at a creative company. I can see myself working in themed entertainment because it requires unique engineering design, creative problem solving, and a heavy use of robotics and automation. This excites me as a way for more people to experience different advancements in robotics and automation as part of their daily lives.

Yes, I would like to become a research scientist in industry now.

I would definitely like to be part of a product-oriented research team in the future, if I join industry. This kind of work is very fulfilling knowing that people use your tools for solving real-world problems.

Yes. Initially, I have been mostly looking for jobs in Robotics companies, but I realized that there are plenty of robotics or control-related positions in other industries as well. I wasn’t even considering the Aerospace industry before my internship, but now, I am looking into a wider spectrum of positions where I can apply my skills and experiences.

The internship gave me the opportunity to experience an industry role for a brief time. The internship influenced me in a positive way since I believe that work life balance is easier to have at Cruise compared to what I see in academia. However, I’m still undecided.

We were thinking of continuing the work as a part-time employee, but unfortunately, Zoox and Penn were unable to reach an agreement due to IP issues.

Yes, I will continue to collaborate with the team past the summer.

I made a few connections with various team leads within the company for my future job search, but I wasn’t able to start any collaborations outside of my internship.

I was mentored by a couple of researchers with whom I plan to continue collaboration.

I focused on my internship during the summer.

I mostly kept my Penn projects on the backburner during my internship. However, I did complete reviewer feedback duties, rewriting some existing papers, and other work.

Since I was out of town for the summer internship, I wasn’t able to do any hands-on fabrication or experiments for my research at Penn. I focused on my internship responsibilities during the workday, but I kept advising an undergraduate student who was working with me during the summer and worked on logistics-planning and theoretical calculations for my research project.

While I was on an internship, I wasn’t officially enrolled at Penn. However, I was still writing a paper which was due at the end of the summer. This paper was a resubmission of a previous paper. Hence, I was able to balance a little bit of research and work. But for the most part, I only focused on the internship.

YES!!!!!

I would recommend my internship to future students who are interested in working for a company that has creatively driven engineering projects.

Yes, Microsoft Research has a good balance between fundamental and applied research, so it’s easy for any PhD student to find relevant problems and collaborators.

My time at ETH Zurich in the Robotic Systems Lab has been important in shaping my future goals. I’m incredibly grateful for having had the opportunity to conduct research alongside such talented and motivated individuals and for the invaluable mentorship I received. It was an impactful summer that I would highly recommend.

100%

Yes. It was a very different experience for me to focus on applying existing robots and systems for another cause instead of developing the robot itself. This allowed me to see a bigger picture for my research here at Penn, and I would highly recommend being able to experience what it is like to be an engineer outside of a research laboratory for gaining the new perspective.

Absolutely! My only disappointment was not getting to meet any of team members in person since the internship was remote. Regardless, I think Cruise is working on very interesting projects and my mentors were very kind and fun.