Q&A with David Esiobu and Aram H. Markosyan from RAISE, Meta AI’s rotational program

April 1st, 2022

This is the final piece of a three-part series that spotlights software engineers participating in Meta AI’s Rotational AI Science & Engineering (RAISE) program, which aims to give participants from diverse backgrounds an opportunity to start their AI journey full-time at Meta. Read more about the RAISE program here.

“AI is one of the most promising frontiers of human discovery. Basic research, precision medicine, transportation, manufacturing — so many aspects of life could be revolutionized by these techniques,” said Software Engineer David Esiobu. “A career in AI seems like one of the best ways to tackle interesting, challenging projects that make a meaningful impact.” Working at Microsoft, Esiobu saw firsthand AI’s power to transform services. He wanted to contribute, but his background was in distributed systems, not AI.

Then he heard about Meta’s new Rotational AI Science and Engineering (RAISE) program, designed for software engineers who, like Esiobu, hope to kick-start a career in AI but lack professional experience in the field. Over 18 months and three rotations, participants — called RAISErs — join world-class research and engineering teams at Meta AI and gain hands-on experience working with a variety of cutting-edge technologies.

In the third and final installment of our RAISE Q&A series, we hear from David and his fellow RAISEr Aram H. Markosyan, both part of the inaugural cohort. David, located in Seattle, came to the program to pursue his interest in the intersection of software, AI, and life sciences. Aram, based in Menlo Park, is leveraging his background in large infrastructure engineering and fundamental research. David and Aram share how they learned about RAISE, how they found themselves at Meta, what they’re working on, and where they see themselves after the program.

If you missed the first two RAISE Q&A, read them here and here.

Applications for the 2022 RAISE program will close on April 15, 2022. For more information and to apply, visit our program page.

Tell us a bit about your career path.

David Esiobu: I discovered programming as a kid when I learned BASIC to make games on a DOS computer, and it’s remained a passion. I went on to study computer engineering at Georgia Tech with a focus on systems programming and bridging the worlds of hardware and software.

Upon graduation, I joined Citrix full-time, where I spent eight formative years. In my last role there I was part of the platform team spearheading the company’s transition to cloud-first offerings. I helped build core services for the effort and developed an interest in distributed systems.

I moved to Seattle in 2016 to work at Amazon, where I learned about building and running services at scale. I worked on Amazon Photos for four years, on everything from clients to back end, and started exploring regression models for sharing suggestions.

Working in healthcare and life sciences runs in my family, and I’d always seen that as a potential calling. When a chance to work on AI-powered services for health care providers at Microsoft came up, I jumped at the opportunity. While there, I worked on back end, deployment infrastructure, and data pipelines supporting researchers.

Aram H. Markosyan My career has oscillated between applied and theoretical science. I was born in Yerevan, Armenia, and I received my first MS in pure mathematics from Yerevan State University. Then I wanted to do more applied science, so I did my second MS in applied mathematics from Université Pierre-et-Marie-Curie in Paris. After that, I earned a PhD from Eindhoven University of Technology, in the Netherlands, in computational plasma physics, which was a good balance of theory and modeling.

That pattern — back and forth between applied and theoretical work — repeated many times in my career. My research has journeyed through optics, chemistry, vacuum science, and computer science. With every iteration, I worked faster and dug deeper into the subject at hand.

During the last months of my PhD, Professor Mark Kushner, the director of the Michigan Institute for Plasma Science and Engineering at the University of Michigan, invited me to work with him — the highest honor. During my two wonderful years at the University of Michigan, I worried about the amount of time it took to make our physics simulation codes work on the high-performance computing cluster. I started to develop a theory of runtime systems, an automatic way of discovering concurrency and parallelism in programs. After several weeks of work, I discovered that this is already a field in computer science! Less than a year later, I took a postdoc position at Sandia National Laboratories, the best place to do research on that topic.

In 2018, I joined Xilinx (now AMD), where I led two large infrastructure projects, one of which was building distributed computing infra from the ground up.

What drew you to AI and the RAISE program?

DE: At Microsoft, my team was pushing the state of the art in speech and natural language processing (NLP) for medicine. I wanted to contribute more in that area, and RAISE is allowing me to make a real-world impact while I deepen my skills in AI. The rotations add a unique opportunity to work in different domains, modalities, and parts of the stack before doubling down in any one specialty.

RAISE is allowing me to make a real-world impact while I deepen my skills in AI.

David Esiobu, RAISE

AHM: I have contributed to several scientific fields, always with the same motivation: to expand our understanding of the world around us. We use cognition in that process, and AI promises to help enhance our cognitive abilities. RAISE is the perfect opportunity to get to know the field from various perspectives.

Now that you’ve completed your first full rotation, how would you describe your experience?

DE: This was my first six months at Meta, and it’s been great learning the culture. I’ve always had to be deliberate about creating work-life balance — I have a tendency to fixate on things until I can see progress, to the detriment of other things in life. The people I’ve worked with here have been respectful of that balance.

Thankfully, my peers are ramping up at the same time. It’s a very social environment at Meta, and we learn faster from each other. Different teams collaborate and coalesce around initiatives in an organic, bottom-up way.

It’s a very social environment at Meta, and we learn faster from each other. Different teams collaborate and coalesce around initiatives in an organic, bottoms-up way.

David Esiobu, RAISE

AHM: Although I’m used to working long hours, it feels like I’ve compressed many years into the past several months. I learned a huge amount about the scale of the problems in AI. My fantastic managers, cohort, and peers have been very supportive, and I’ve met so many super-talented Metamates.

What are you working on?

DE: I’m currently working with the translation team on rewriting the language identification model and expanding the number of languages and dialects it supports. That’s been a great introduction to the experimentation process — I’m coming to appreciate how critically important high-quality data is.

AHM: I’m working on improving the robustness of large-scale transformer models used in NLP and computer vision.

Have any skills from your previous career been particularly useful as you ramp up?

DE: During my first rotation with the AI Foundation team, my experience with back-end architecture and distributed systems helped in building observability tools for distributed training infrastructure.

AHM: My mathematics background was a tremendous help in reading AI papers, and my software engineering past helped me ramp up quickly on infrastructure.

What are your plans after RAISE?

DE: Language seems to be fundamental to the way we think and understand — and possibly to how virtual agents do as well. I hope to continue in NLP and natural language understanding, eventually bridging back to the life sciences. NLP techniques have already shown promising results in fields like genomics, and I’d like to be part of that.

AHM: I will feel privileged to continue doing research at Meta after RAISE. The problems we are facing in terms of scale and complexity are at the forefront of state-of-the-art AI research.

What advice would you pass along to someone thinking of applying to RAISE?

DE: Be curious. Many topics at first appear difficult to grasp, but approaching with curiosity opens you to learning quickly and trying new things.

The problems we are facing in terms of scale and complexity are at the forefront of state-of-the-art AI research.

Aram H. Markosyan, RAISE

Also, be deliberate about what you want to do. The field is broad. Within Meta, things move quickly, and it’s easy to lose the signal in the flow of information. Having a clear direction will help you push through challenges and discern what to pay attention to.

AHM: Don’t expect a smooth transition. There will be a lot of learning — mostly on your own time — as well as frustration and tight timelines. But your fellow RAISErs, the RAISErs from the past cohorts, your managers, and your peers will be there to champion you.