Amanda Dsouza is from sunny Goa, India and is a full-time OMSCS Student. Before joining OMSCS, Amanda was a Lead Data Scientist at Fractal Analytics where she led a team in natural language processing and text mining problems. Amanda was inspired to join OMSCS when she changed career paths from software development to data science a few years ago and she was looking to strengthen her foundation in some key areas. As for why OMSCS, she “found that the program was high quality, and after doing some research and speaking with students who had graduated from the program, I decided it was a great fit.” Besides the quality of the course content, Amanda found “the peer and teaching network and the tools we use interacting as a class to be phenomenal.” She is also impressed that new relevant courses keep getting added to the program.
"The [OMSCS] program was high quality, and after doing some research and speaking with students who had graduated from the program, I decided it was a great fit."
Amanda is currently working with two other researchers on a problem in adversarial imitation learning. Their goal is to improve an information bottleneck which, when applied to an adversarial imitation learner, learns a policy that not only performs well on the original environment, but also generalizes better to unseen conditions. Amanda and her two colleagues aim to show that their bottleneck learns a more accurate latent representation, leading to more stable policies. Amanda contends that she started thinking about this idea earlier in 2022 and has since received important feedback on the work from other researchers. Furthermore, as Amanda astutely explained, “Georgia Tech has a strong research practice and one that I have been trying to take advantage of.”
Besides Computer Science, Amanda would one day like to learn more about the fields that influence AI and that which influence Reinforcement Learning, specifically neuroscience and psychology. Amanda brilliantly contends that “while much of our work today is based on rational healthy agents, which has been essential for making progress in the field, only recently have we increased our efforts in tackling irrational agents. I think learning more about neuroscience and psychology will be essential in developing and deploying agents in the real world.” Well said, Amanda!
"I'm impressed with the challenges the courses presented, their focus on reading research papers, topic discussions by peers, and the involvement of professors and TAs."
There are two courses that Amanda has taken thus far which she crowns her favorites: CS-7642 Reinforcement Learning and CS-7643 Deep Learning. Reinforcement Learning introduced her to the subject matter, which she previously did not know much about, and both courses led her to her research project. Amanda explained that she is “impressed with the challenges the courses presented, their focus on reading research papers, topic discussions by peers, and the involvement of professors and TAs.” Her praise does not end there; in fact, she continues, “to be honest, I’d love to work on research with some of Georgia Tech’s teaching staff.”
So, what does Amanda do when not working on inspiring research and taking OMSCS courses? She usually sketches or goes on short bike rides to clear her head, after a long day of watching lectures or working on problems.
As for what is in Amanda’s future after graduation, though it has been great to focus solely on the program and on her research this year, she is eager to get back to industry doing research or working on a product in Machine Learning. But it won’t be the end of her OMSCS journey, as she plans to continue taking courses after she graduates.