AI Research Summer Fellowship Program

graphic of the world and a brain being connected by people and AI

Organized by the Stevens Institute for Artificial Intelligence (SIAI), the AIRS Fellowship Program empowers selected Stevens undergraduates to embark on an exciting research journey on a topic of their choice.

The SIAI's 2021 AIRS Fellowship Program

Humans and AI: An Evolving Partnership

The theme for the 2021 AIRS program was "Humans and AI: An Evolving Partnership". AI is already having an impact across myriad aspects of society, from mobility to healthcare, from education to finance, and more. In many of these scenarios, we find humans at the center: humans getting from place to place, humans striving toward wellness, humans teaching, humans learning, etc. As a scholarly community, we are now beginning to ask questions like, “How should AI augment human capabilities?” rather than, “When will AI replace humans?” This relationship between humans and AI is evolving. Selected projects in the 2021 AIRS program targeted questions under this theme to better understand and enhance the relationship between humans and AI.The AIRS Fellowship Program solicits independent undergraduate research project proposals from Stevens Undergraduates.Due to recent changes in the institute, the AIRS Fellowship Program will not be offered for Summer 2022.  We look forward to launching more opportunities through the SIAI during the 2022-2023 academic year, so please watch your emails and the website for updates.

2021 AIRS Fellows and Projects

MODELING HOMELESS CHARACTERISTICS TO SUPPORT INTERVENTIONS

Jared Donnelly (Computer Science, SES, 1st Year) and Jolene Ciccarone (Software Engineering SSE, 1st Year)

Mentor: Professor Samantha Kleinberg

Abstract: Homelessness is a widespread issue that has a ripple effect across communities, affecting everyone from the victims of homelessness to the communities struggling to help them.

The lifespan of a homeless person is 30 years below the average, and the mean lifespan of someone living on the streets is only 11 years. With more homeless people in NYC than the entire population of Hoboken, it is clear that there is a homelessness problem. As a result, nonprofits like Built for Zero and Project Renewal have put it upon themselves to help homeless populations through healthcare, housing, and jobs. Since their efforts have had significant positive impacts on homelessness, especially with veteran, youth, and chronic homelessness, we would like to build upon their work. We aim to approach homelessness from a preventative rather than a reactive perspective, which is highly beneficial in the long run, especially since responses to pressing issues tend to have a history of long delays and poor implementation. Our goal is to utilize AI to provide future predictions of homeless populations and their characteristics for non-profits, government agencies, and policymakers so that they can make informed decisions and preparations.


IMPACT OF AI COMPANION ON NURSING-HOME RESIDENTS

Sakina Rizvi (Business and Technology, BUS, 3rd Year)

Mentor: Dr. D. N. Lombardi (Stevens Healthcare Educational Partnership)

Abstract: The goal of my research is to gain a comprehensive understanding of the emerging role of AI in healthcare, with a specific focus on identifying programs, opportunities and positive impacts on aging facility residents derived from individual interaction with a robot. Since the objective for this program is to better understand and enhance the relationship between humans and AI, I plan to specifically research AI in nursing homes contending with the phenomenon of aging isolation, with an intent of ascertaining if structured robot interaction can be a solution strategy.

My research will include the exploration of possible challenges and ethical concerns in order to maximize the benefits associated with AI technologies. A field survey of existing AI applications in the aging services sector, as well as the potential for improving the safety, quality, and efficiency of healthcare through robotics will also be explored.

While social isolation has been considered a dangerous health risk for older Americans, COVID- 19 intensified the dynamic in the past year. By the end of my research project, I will have successfully analyzed how social robotic interaction can represent a potential solution to the social isolation of aging facility residents.


NATURAL LANGUAGE PROCESSING AI IN CUSTOMER FEEDBACK ANALYSIS

Pawan Perera (Electrical Engineering, SES, 2nd Year)

Mentor: Jia Xu 

Abstract: Natural language processing (NLP) is a subsection of AI that handles human language data, specifically recognizing and processing it. A specialized denomination of NLP, that is considered an AI-hard problem, is natural language understanding (NLU), which expands upon NLP by being able to infer, summarize, and comprehend human language data. As of yet, there have been few practical applications of NLU but its potential capabilities are of great significance. This study researches the development of an NLU to allow customers and manufacturers to quickly extract relevant key insights from large amounts of customer feedback and reviews on a product sold online. Prior research on NLUs has illustrated several common components that can be sourced from these studies. However, semantic theory will be the main hurdle of this study since it is necessary to establish the method in which the system interprets the data, judges relevance, and comprehends overall. In turn, research will aim to tailor the semantic theory to meet our needs here. This research on NLU development to analyze customer feedback has profound implications as one of the first practical applications of this AI-hard problem, which can pave the way for future use of this evidently revolutionary AI development.


ALPHAAI: NOT ALONE INVESTING

Ryan Finegan (Business and Technology, BUS, 3rd Year)

Mentor: Dragos Bozdog

Abstract: AlphaAI is a project targeted toward inexperienced investors during this new age of commission-free trading with the purpose to decrease wealth disparities through providing analytics and suggestions utilizing artificial intelligence. Recently, there has been an influx of new market participants as retail investors flood into trading platforms that are more user friendly and affordable. This progressive development has been met with obstacles, such as misguided advice delivered to oblivious market entrants regarding volatile or risky securities. This project seeks to apply machine learning models to aid in investors' short- and long-term investment decisions through time series forecasting, security sentiment analysis, and equity screening and recommendations. This software pursues analysis through various techniques such as clustering, ensemble methods, recurrent neural nets, and natural language processing to offer an individual investor the tools to increase investment returns and alpha. Investor intuition and AlphaAI's diverse set of models work cooperatively towards the goal of steadily increasing one's wealth through consistent, informed investment decisions. AlphaAI is meant to be a resource available to accompany green investors while navigating the ever changing and challenging equity markets.


SKINCARE AND AI: HOW AI TECHNOLOGY CAN DECOMMERCIALIZE LUXURY SKINCARE SERVICES

Serena Lee (Software Engineering, SSE, 2nd Year)

Mentor: Mukund Iyengar

Abstract: Millions of Americans struggle to afford health insurance and it is difficult to find a suited doctor for their needs. Many of these people are teenagers suffering from acne-related conditions and can not see a medical professional. This project aims to create a program that will allow teens to “self-diagnose” themselves of their acne-related problems without the costs of seeing a dermatologist and receiving equal results. By using convolutional neural networks (CNN), we can create a system that can find similarities between data sets and apply them to real-life situations. Teenagers around the country can figure out their skin conditions without spending a penny and from the comforts of their home.


VISUALLY ENHANCED PODCASTS

Burak Yesil (Computer Science, SES, 1st Year)

Mentor: Jason Corso

Abstract: Podcasts are an efficient way for people to gain insight into other people’s views and current events. However, as listeners may not be familiar with a specific person or subject being discussed, they may have a hard time following the conversation and will be discouraged from listening to future podcasts on the subject. My AI model will solve this issue by providing people with basic knowledge by automatically displaying pictures and article links, which will be generated by analyzing keywords from podcasts, to help listeners get a better understanding of what is being discussed.

In my research, I will use Google AI’s TensorFlow platform and Facebook’s PyTorch Platform to obtain the data sets needed to train my model in the Jupyter Notebook Environment. Some specific datasets include but aren’t limited to, “CelebA” (a celebrity faces data set) and “VoxCeleb” (a large-scale audio-visual dataset of human speech). Through the use of Back Propagation, I will lower my AI model’s margin of error. One major application of my AI model will be to visually augment podcasts, allowing listeners to be more engaged and knowledgeable, enhancing their overall experience. This model can be used by platforms such as Spotify.

Frequently Asked Questions