December 2, 2020

How Facebook Core Systems’ Shruti Padmanabha transitioned from PhD candidate to Research Scientist

By: Meta Research

Our expert teams of Facebook scientists and engineers work quickly and collaboratively to solve some of the world’s most complex technology challenges. Many researchers join Facebook from top research institutions around the world, often maintaining their academic connections through industry collaborations, partnerships, workshops, and other programs.

While some researchers come to work at Facebook after extensive careers in academia, others make the transition toward the beginning of their career. Shruti Padmanabha, who joined Facebook after receiving her PhD in computer science and engineering, is one such example.

Padmanabha is a Research Scientist on the Disaster Recovery team within Facebook Core Systems. Her focus is on building distributed systems that are reliable and tolerant to failures in their hardware and service dependency stack. Padmanabha’s journey with Facebook Core Systems began with a PhD internship, an experience that encouraged her to switch fields from computer architecture to distributed systems.

We sat down with Padmanabha to learn more about her experience joining Facebook full-time directly after earning her PhD, as well as her current research projects, the differences between industry and PhD research, and her academic community engagement efforts. Padmanabha also offers advice for PhDs looking to transition to industry after graduation.

Q: Tell us about your experience in academia before joining Facebook.

Shruti Padmanabha: After earning an undergraduate degree in electrical engineering in India, I moved to Michigan to pursue my master’s and PhD in computer science and engineering at the University of Michigan, in Ann Arbor. I was interested in how low-level circuits came together to form computers, and started working research problems in computer architecture. Specifically, I focused on energy-efficient general-purpose architectures (like those that drive desktops).

What excited me about grad school was the opportunity to continuously extend state-of-the-art technology. There, I learned the importance of communication, both written and in person, and to navigate myself in a male-dominated environment (I was the first and only woman in my adviser’s lab!) by leaning on and volunteering in underrepresented groups for grad students. I also found passion in teaching and mentorship activities.

But foremost, I valued the company of the brightest peers and network that being in grad school brought me. It was a proud feeling to be treated as a peer by professors that I greatly respected!

Q: What has your journey with Facebook been like so far?

SP: My first experience at Facebook was a 2015 summer internship in Core Systems. This internship was a leap for me since my expertise was in computer architecture, not in distributed systems. I worked on a proposal to improve energy efficiency of Facebook data centers by dynamically scaling machine tiers based on traffic patterns. The insights we gained from this early experimental work influenced the design of some of our production capacity management systems today.

During my internship, I had the opportunity to work on challenging problems for state-of-the-art technologies and to work with smart, driven peers. I liked that I could innovate on problems with more immediate, real-world applications than in academia. This is why I decided to pursue a full-time position at Facebook after 27 years in school.

When I joined, I was given the choice of being a computer architect on another Facebook team or to switch fields and rejoin Core Systems. I had a very positive internship experience on the Core Systems team, and I liked that they maintained strong academic collaborations, so I decided to permanently switch fields and move to distributed systems.

My first position was on an exploratory team that was trying to think of novel ways to distribute Facebook systems in data centers of different sizes and in different geological locations, which gave me a crash course in real-world distributed systems of all kinds. It also provided me with the unique opportunity to learn about the physical side of designing data centers at Facebook’s scale, from power transformers to submarine network backbones to CDNs.

I’ve learned that there’s always room to learn new things and grow within Core Systems. To keep a healthy flow of fresh information flowing, I lead biweekly brown bag sessions where teams present technical achievements and design challenges, as well as help present introductory infrastructure classes to Facebook n00bs.

Q: What are you currently working on?

SP: My current focus is on improving fault tolerance and reliability at Facebook. Facebook’s service infrastructure consists of a complex web of hundreds of interconnected microservices that change dynamically throughout the day. These services run on 11 data centers, which are designed in house and located across the globe.

At this scale, failures are bound to happen — like a single host losing power, a power transformer getting hit by lightning, a hurricane posing a risk to an entire data center, a single code change leading to a cascading failure, and so on. Our team’s focus is to help design Facebook’s services in a way that such failures are tolerated gracefully and transparently to the user.

Our OSDI paper from 2018 talks about one mitigation approach of draining traffic away from failing data centers. Justin Meza and I also gave a talk at Systems@Scale 2019 where we described the problem of handling power outages at a sub–data center scale. The solution required building solutions that span the stack — from respreading hardware resources across the data center floor to balancing services across them.

Q: What are some of the ways you’ve shown up in the academic community?

SP: I’ve stayed in touch with the academic community by attending conferences, participating in program committees, and mentoring interns. I served on the program committees for a few academic conferences (ISCA, HPCA), as well as for Grace Hopper. I enjoy participating in events where I can reach out to PhD candidates directly, especially those from underrepresented groups. For instance, I moderated a panel at MICRO 2020 on Tips and Strategies for PhD candidates looking for jobs in industry, and participated in the Research@ panel in GHC 2017.

At Facebook Core Systems, we maintain and encourage a strong tie to academia, in terms of both publishing at top systems conferences and awarding research grants to academic researchers, for which I also had the opportunity to help out.

Q: What are some main differences between doing PhD research and doing industry research?

SP: In a PhD candidacy, one’s work tends to be fairly independent and self-driven. In industry, on the other hand, projects are distributed across team members who are equally invested in its success. I had to deliberately change my academic mindset of being solely responsible for the delivery of a project to working in a collaborative environment as an individual contributor.

Industry research also might have the advantage of mature infrastructure and access to real-world data. To me, not having to build/hack simulators and micro-benchmarks was a breath of fresh air. At Core Systems, development progresses from a proof of concept to being tested and rolled out in production relatively quickly.

I’ve also observed a difference in how projects are prioritized in industry and how success is measured. Within Core Systems, we have the freedom to choose the most important problems that need to be worked on within the scope of company-wide goals and build long-term visions for their solutions. Half-yearly reviews are the checkpoints for measuring success toward this vision.

Q: For computer science PhDs curious about transitioning to industry after completing their dissertation, where would you recommend they start?

SP: Explore different sides of the industry through internships, and leverage your professional networks. We’re fortunate that, in our field, internships are usually plentiful. Experiment with different kinds of internships at research labs and product groups if you can. These opportunities will not only give you a window into the kind of career you could build in the industry but also help you build connections with industry researchers. Many industry jobs and internships have rather generic-sounding descriptions that make it hard to envision the work involved, so it’s best to get hands-on experience.

Also, leverage connections through your adviser and alumni network, and engage with industry researchers at conferences. Be sure to discuss the breadth of research at their company and their job interview processes. Lastly, don’t be afraid to step outside of your research comfort zone to try something new!