Shereen Oraby graduated from the Baskin School of Engineering with a Ph.D. in computer science in 2019. Her academic focus on natural language generation and the Silicon Valley internships she participated in as a student have led her to a career as an applied scientist with Amazon Alexa AI.
Oraby first became interested in natural language processing (NLP) when she was a student in Egypt, where her family is originally from. She studied computer engineering for her bachelor’s degree, but she also had an interest in language, so NLP seemed like a natural fit.
“I’d read some articles on interesting research going around in NLP, and specifically for Arabic, and I found that that was kind of a nice hybrid between computers and language,” Oraby says.
Oraby began doing work on Arabic NLP as a Master’s student in Egypt. After her Master’s, her desire to be on the cutting edge of NLP research brought her to UCSC, where she was a member of Computer Science Professor Marilyn Walker’s Natural Language and Dialogue Systems (NLDS) lab.
Cultivating Diversity in NLP
Having worked on both English and Arabic NLP, Oraby perhaps understands better than most the importance of having diversity in the field.
“In order to understand how different languages work, you need people with different backgrounds. Having experience with different languages helps you understand what makes each of them unique,” she says. “Having lots of different perspectives is really good to get the big picture for NLP.”
Part of what she appreciated about being at UCSC was that Professor Walker made a conscious effort to diversify her lab. “My advisor was always super supportive and a really easy person to talk to, and she tried to make sure that our lab was really diverse and we had different kinds of people working on different kinds of projects and I always really appreciated that,” she says.
“I think [Natural Language Processing] is an up-and-coming field, and it will continue to be because there are a lot of unsolved problems in NLP. It is a good time to be working on it, and UC Santa Cruz was a good place to be working on it.”
Working with Industry
Oraby had three different internships while she was a student at Baskin Engineering: one with IBM, another with Megagon Labs, and finally one with Amazon. She says that the University’s proximity to the Silicon Valley facilitated her ability to get these internships and the hands-on industry training they provided.
“We would always work on projects that were really relevant, not just in the research community, but also in industry,” she says. “We’re so close to all of these different companies, and people would come in for talks, sometimes with the intention of recruiting people or trying to see if there’s a fit for collaboration with our lab. That was super helpful and most of my internships came about in that way.”
Another way that Oraby made connections with industry as a student was by participating in the very first Alexa Prize competition, in which teams throughout the nation compete to build a social bot that someone could have a conversation with for 20 minutes about any topic in the world.
“That’s a hugely fanciful thing,” Oraby admits, “but the idea is to take gradual steps.”
At the end of the competition, the bots compete against one another and are judged on how engaging they are. UCSC was one of only a handful of universities that was selected to be funded in the inaugural competition three years ago, and it has had a team every year since, including this year.
“It’s nice because you get resources, funding, and some good perks, and it’s nice to be doing work that is clearly industry applicable,” she says.
A new research direction
In addition to providing her with tools that she could use later in industry, working on the Alexa Prize also influenced Oraby’s academic focus.
“We got involved with the Alexa prize competition, and that kind of got me interested in some of the newer hot topics in NLP, which were things like language generation,” she says.
Oraby’s initial research had focused on the challenging task of teaching machines to detect sarcasm in social media text.
“It’s one of those topics that people are always working on, because sometimes it’s hard even for a human to evaluate if something is sarcastic or not, so it’s very hard for a machine learning model to tell,” she says.
After becoming interested in language generation during the Alexa competition, Oraby maintained her focus on language styles, but pivoted to studying how to generate them rather than detect them. In her doctoral thesis, she focused on how to train a dialogue system, such as a personal assistant, to vary the style of its responses to be, for example, more verbose or succinct or descriptive. This type of stylistic variation will allow machines to be more human-like and conversational in their interactions with people.
Oraby’s research, combined with her internships and industry connections, helped her to land a job at Amazon right out of graduate school. Today, she works as an applied scientist in the Amazon Alexa AI organization in the Dialog Systems team. Her team uses their knowledge of NLP and dialog systems to build machine learning frameworks that developers can use to rapidly create Alexa skills.
“What I like about it is that it is an applied role, which means that the goal of the research is directly applied into a product,” she says. “Whatever you do, the goal is to improve on a product, improve on a customer experience.”
Advice for students
For students who are currently looking for careers in industry, Oraby says that internships are crucial, and to get them, she advises reaching out to people at conferences and talks. When that is not possible, such as under the current COVID-19 restrictions, she says that emailing authors of research papers you are interested in can also help you make great connections.
“You might not realize how what you’re working on is really relevant to somebody, but it could turn out to be very applicable,” she says.
Oraby admits that putting yourself out there can be difficult, and she is grateful that her NLDS lab made it easier through strong peer support and conferences and other networking events. One of the best things about studying at UCSC, according to Oraby, was the connections she was able to make not only with industry professionals, but also her fellow students.
“We had a big lab, around 10 to 12 people, and when we would often have six or seven people going to the same conference,” Oraby says. “People knew that the UCSC NLDS lab would go to conferences in a big group, and that was really fun. It’s nice to have a good set of friends that you’re working with, because grad school is a really big journey. It was a good experience for me.”