2019 NCWIT Collegiate Award Recipients

The NCWIT Collegiate Award honors the outstanding computing accomplishments of undergraduate and graduate women. Conferred annually, the award recognizes technical contributions to projects that demonstrate a high level of innovation and potential impact.

To learn about the award recipients' contributions, please view the descriptions below.

Winners

• Isabel Gallegos, Stanford University, Unsupervised Learning Methods to Perform Material Identification Tasks on Spectral Computed Tomography Data
  This project developed an unsupervised machine learning algorithm to differentiate materials with very similar compositions using reconstructed high-energy spectral computed tomography (CT) data. It improves upon existing methods by pre-processing the data set to remove unwanted noise and artifacts, then uses a new algorithm called iterative hierarchical clustering to classify data into a prespecified number of subsets. In tests using ceramic compounds, the new method was the only algorithm to classify the dataset with 100 percent accuracy. (View the project online.)

• Sharon Lin, Massachusetts Institute of Technology, Reverse-Engineering Peripheral Drivers for Abstract Modeling
  This project involved reverse engineering assembly code from firmware for IoT connected hardware devices in order to parse for drivers being used in running the device, making it easier to determine functionality and identify vulnerabilities. This is accomplished via the creation of a generalizable script that is able to take the lowest-level information about bits being translated from registers to virtual memory, as well as low-level operations, making it possible to extrapolate the higher-level code running on the device. (View the project online.)

• Tayebeh Bareini, Wayne State University, Resource Management Mechanisms in Edge Computing Systems
  This project addresses the growing computation needs of increasingly sophisticated mobile technologies by designing and implementing efficient resource management mechanisms for mobile edge computing (MEC) systems. This new approach optimizes efficiency through multiple innovations, including a two-phase algorithm for determining the best match between application and server components; an energy-aware mechanism for provisioning of edge resources; and an envy-free, auction-based method for pricing edge resources. (View the project online.)

• Annika Muehlbradt, University of Colorado - Boulder, Exploring the Design of Audio-Kinetic Graphics for Education
  In this project, existing technology - a pen plotter tool - is repurposed as a way for teachers to communicate spatial information, such as diagrams on a whiteboard, in a tactile format for visually impaired students. A teacher authors audio-kinetic graphics using a stylus on a touchscreen device. The system records the stylus movements as x, y positions and timestamps at millisecond intervals. On playback by the student, the software synchronizes the audio content with movements of the pen plotter. (View the project online.)

• Katherine Spoon, Indiana University - Bloomington, Dytective: Automated Early Detection of Dyslexia using Neural Networks
  This project employs a machine learning approach to dramatically improve early diagnosis rates for dyslexia. The researchers developed a tool that uses a multi-stream deep convolutional neural network to analyze handwriting samples and learn which characteristics are likely to correlate with dyslexia. This helps educators to identify and support students who might have this learning disability much sooner, by getting students into the queue to be properly diagnosed by trained professionals by the time they leave elementary school. (View the project online.)

• Courtney Thurston, Embry-Riddle Aeronautical University - Daytona Beach, Creating a Fair Learning Platform for Computational and Data-Enabled Science and Engineering Instruction
  The aim of this project is to increase access to Computational and Data-Enabled Science and Engineering (CDSE) courses while reducing the effects of bias in assessment through a new learning management system (LMS) designed for the unique needs of educators and students in this field. The platform supports collaboration across institutions, enabling small departments to pool their resources, while using machine learning to make fair assessments of individual students’ effort and contribution to group assignments.

Honorable Mentions

• Samsara Counts, George Washington University, Deep Learning Tools to Improve Eating Disorder Recovery
  Recognizing that exposure to triggering online content can be a cause of setbacks for people recovering from an eating disorder (ED), this project developed machine learning tools to identify pro-ED websites. The application combines a convolutional neural network that detects ED images with two software tools that assess websites for ED content: one for diagnostics (designed to be used by clinicians), the other for filtering out triggering material (to be used by individuals in recovery). (View the project online.)

• Anna Dodson, Dartmouth College, The Computation of Perception: Novel Computational Approaches to Improve MPEG Compression
  This project draws on the neuroscience of visual processing to improve the methods used to compress large video files so that the resulting images more accurately align with the ways in which the human brain filters data based on spatial mappings. Extending earlier research that applied Riemmanian geometric principles to compression of still image files, this project introduces a new encoding and decoding system that compresses videos to a smaller size while maintaining the same perceived image. (View the project online.)

• Amel Hassan, Tufts University, Augmented Reality Framework for Collaborative Service Robots
  As the presence of robots in human-populated environments expands, the field of Human-Robot Interactions (HRI) is working to increase understanding and communication between humans and robots to improve task achievement and reduce the potential for injury. The application developed in this project allows users to better conceptualize what a robot is “thinking” by using augmented reality to create a visual representation of the robot’s sensory inputs and decision-making processes. (View the project online.)

• Amber Johnson, Purdue University, Population Scale Longitudinal Data Analytics in Healthcare
  This project makes it easier for health care providers to quickly extract useful information from large quantities of patient data stored in multiple formats by performing statistical analyses and creating visual representations of the results. These models give doctors baseline values of a patient’s current state and probabilistic suggestions about how their state could change given certain treatments. This software will enable practitioners in rapid-response settings, such as ICUs, to make fast, data-driven care decisions. (View the project online.)

• Veenadhari Kollipara, University of South Florida - Tampa, Empowering Farmers with a Sustainable Precision Agriculture End-to-End Solution
  This project supports farmers in adopting high-efficiency Precision Agriculture methods by designing lower-cost, easier-to-use equipment and software platforms. The drone-enabled system carries a soil-sampler payload with onboard Arduino sensors, IR camera, GPS and a sprayer system, while a “farmer-friendly” dashboard enables visualization of field data so that farmers can make better and more informed decisions regarding planting and crop care, reducing wastage of costly inputs like water, fertilizers, and pesticides. (View the project online.)

• Ashlie Martinez, University of Washington - Seattle, Bounded Black-Box Crash Testing
  This project developed a testing technique for file systems to identify crash-consistency bugs, which are bugs in file system code that cause data loss or corruption when a computer crashes during a file-system operation, but have no effect if no crash occurs, making them difficult to detect outside of a crash. This new tool safely generates crash states while recording file data in order to determine if any data is lost or corrupted during the crash. (View the project online.)

• Vilina Mehta, Stanford University, Discovery of a Novel MicroRNA-Targeted Approach to Overcoming Drug Resistance and Advancing Therapeutic Strategies for Glioblastoma Multiforme
  This project uses computational modeling to explore and test a new approach to treatment of glioblastoma brain tumors. After discovering a small molecular RNA inhibitor molecule through molecular modeling software simulations, the researcher was able to construct a new treatment approach based on a sequential release of microRNA inhibitors which modifies glioma cells from a drug-resistant state to drug-sensitive state, thereby increasing the cancer’s responsiveness to current chemotherapy drugs.

• Taylor Miller-Ensminger, Loyola University Chicago, The Bacteriophages of the Urinary Microbiome
  This project utilized multiple data computation tools, including VirSorter and others, to conduct the first-ever study cataloging the population of bacteriophages existing within the recently-discovered microbiome of the human bladder and urinary tract. This study uncovered 457 phage sequences, most of which were novel, exhibiting no discernible sequence homology to public data repositories. Similar phages were, however, found within the microbiota of different women, suggesting a core community of phages within the bladder. (View the project online.)

• Samhita Pendyal, College of WIlliam and Mary, The Effect of Clustering Terms on Improving Literature Based Discovery
  This project explored a variety of methods to determine whether applying unsupervised machine learning clustering techniques could be used to improve the performance of literature based discovery (LBD) in natural language processing applications. This study used clustering of terms to more accurately reflect the ranking of a concept and more readily identify associations between concepts. A two-sided T-test was used to conclude that clustering does, in fact, improve performance. (View the project online.)

• Swetha Prabakaran, University of California - Berkeley, Teho
  This project seeks to improve upon existing stress reduction technologies by incorporating machine learning. Teho is a new web application and calendar integration program that prompts users to enter data about their mental or emotional state after completing various scheduled activities. As it learns which activities are likely to led to states of high or low stress, it can them offer scheduling recommendations to help balance more stressful activities with more relaxing ones.

• Abby Stylianou, Washington University in St. Louis, TraffickCam: Image Analysis to Combat Human Trafficking
  This project aims to combat human trafficking by helping investigators use photographic evidence to identify the specific location in which a trafficking victim was photographed. The project includes a database of several million photographs of hotel rooms, collected from publicly available travel websites and submitted by users of a mobile application. TraffickCam then uses deep convolutional neural networks to support image-based search for members of law enforcement. (View the project online.)

• Angelique Taylor, University of California - San Diego, Robot Perception of Human Groups in Real World Environments
  This project develops an algorithm for unsupervised machine learning that allows robots to predict and respond to the ways humans move through an environment when in groups, as opposed to individually, thereby increasing public safety in situations where humans and robots interact. The researcher created a process by which a robot can identify pedestrians and groups, then analyze their movements by using convolutional neural networks to estimate the similarity of their placement from frame to frame. (View the project online.)