Center for Undergraduate Research in Viterbi Engineering (CURVE) Fellowship

• Principal Investigator: Jennifer Treweek
• Website: sites.usc.edu/treweek-lab/
• Projects
  • 3D imaging and analysis of the intact rodent eye
    • Project Description: The goal of this project is to trace cellular connectivity from the retina-to-brain in intact tissue specimens. This project entails the dissection, tissue-clearing, labeling, and imaging of the rodent eye, optic nerve, and visual centers in the brain.
    • Open to: CURVE First-Time Researchers,CURVE Continuing Researchers
    • Student Responsibilities: Wetlab work: tissue dissection, hydrogel-embedding, histology (antibody labeling), tissue-clearing, and confocal imaging
    • Drylab work: computational image analysis of the retina (cell counting in different retinal layers, quantification of retinal layer thickness, tract tracing, etc)
    • Preferred Majors: Biomedical Engineering
    • Preferred Skillsets: basic wetlab skills (pipetting, calculating dilutions, making stock solutions, use of a chemical fume hood)
    • desired but not required: working with tissue specimens (e.g., histology, light microscopy)

• Optimizing viral vectors for gene delivery to the rodent brain
  • Project Description: The goal of this project is to achieve widespread, brainwide expression of neuronal activity sensors through the use of blood-brain-barrier (BBB)-crossing adeno-associated virus (rAAV) vectors. This project entails producing high-quality, high-titer viral vectors using triple transfection in mammalian cell culture. In addition, this project could include a drylab component: i.e., the creation of an electronic database that document the viral vectors which achieve experimental success - based on a set of evaluation parameters.
  • Open to: CURVE First-Time Researchers
  • Student Responsibilities:
    • Wetlab student: DNA plasmid production and purification, mammalian cell culture, virus purification via density gradient columns and centrifugation, histology of rAAV-transduced brain tissues (antibody-labeling, slide-mounting, confocal imaging)
    • Drylab student: creation of a website to host lab data on experimental parameters and AAV success metrics; also computational image analysis for counting AAV-transduced cells in brain regions-of-interest
  • Preferred Majors: Biomedical Engineering
  • Preferred Skillsets: Preferred skills/experiences are related to tasks in job description:
    • Wetlab student: basic wetlab skills (e.g., pipetting, preparing dilutions and stock solutions, working in a bisafety cabinet and/or knowledge of celll culture techniques would be ideal); knowledge of basic histology "best-practices" would also be ideal
    • Drylab student: basic computer skills (website creation, methods for tabulating datasets, etc); basic analytical skills; for computational image analysis, skills related to either automated/semi-automated methods for cell counting and image analysis - e.g., using imageJ/fiji)

• • Principal Investigator: Bo Jin
  • Website: https://composites.usc.edu
  • Projects
    • Additive Manufacturing of Structures with High Compressive Strength
      • Project Description: We are designing structures for high compressive strength, and they are all 3D printed. The structure should reach/beyond 10k lb. compressive strength and will need to be as tall/light as possible. We will support multiple advanced 3D printers with large print-bed and bed-heating functions for shape/warping compensations, and unlimited printing materials for the project. Satisfying structures will attend the annual SAMPE AMC additive manufacturing contest from which we have won a national 3rd in the past. Great honor and a positive plus to student’s resume.
      • Open to: CURVE First-Time Researchers
      • Student Responsibilities: With the help of faculty and veteran students, conduct meetings and the design of your structures. Design the CAD models using whichever software you are familiar with or trained by veteran students. Print the structures using the lab printers and maintain successful printing processes.
      • Preferred Majors: Aerospace & Mechanical Engineering,Astronautical Engineering,Civil & Environmental Engineering,Industrial & Systems Engineering
      • Preferred Skillsets: No prerequisite skills required. In-person lab participation required.
  • Manufacturing, Processing, Testing, and Machine Learning Predictions of Composite Materials Voids and Structural Strength
    • Project Description: This project focuses on employing machine learning algorithms to detect voids in aerospace composite materials, utilizing data from micro-CT imaging to predict material and structural performance for aerospace and automotive parts.
    • Open to: CURVE First-Time Researchers
    • Student Responsibilities: Work with external industrial advisors and conduct manufacturing, processing, and testing of composite materials and structures, and work with graduate students from USC CS department to apply machine learning algorithms developed to predict material void
    • Preferred Majors: Aerospace & Mechanical Engineering, Astronautical Engineering, Civil & Environmental Engineering, Computer Science, Electrical & Computer Engineering, Industrial & Systems Engineering

• Principal Investigator: Yong Chen
• Website: https://viterbi-web.usc.edu/~yongchen/
• Projects:
  • 3D Printing Research
    • Project Description: Participate in novel 3D printing process, material, and machine development and work with Ph.D. students.
    • Open to: CURVE First-Time Researchers
    • Student Responsibilities:
      • Help to design prototype components
      • Perform experiments and analyze data
    • Preferred Majors: Aerospace & Mechanical Engineering
    • Preferred Skillsets: Students are desired to be interested in hands-on work.

• Principal Investigator: Geoffrey Spedding
• Website: https://drydenwt.usc.edu/
• Projects
  • Aerodynamics of wings at low Reynolds number
    • Project Description: Establish separation lines and points of flow control for a small wing. Three-dimensional descriptions and control will be required
    • Open to: CURVE First-Time Researchers
    • Student Responsibilities: Learn instrumentation and how to run a sweep through incidence angles while taking and analysing force coefficient data.
    • Preferred Majors: Aerospace Engineering
    • Preferred Skillsets: Matlab, basic Aero, such as from AME 105

• Principal Investigator: Maral Mousavi
• Research Website: https://sites.usc.edu/mousavi/
• Projects
  • This project aims to design and develop a highly sensitive biosensor for detecting calcium deficiency, a condition that can lead to osteoporosis and other health complications. The biosensor will utilize potentiometric techniques to selectively and accurately measure calcium ion concentrations in biological samples. By incorporating a calcium-selective membrane and optimizing sensor response, this device will enable early diagnosis and monitoring of calcium deficiency in individuals. This rapid, non-invasive tool has the potential to improve healthcare outcomes by providing accessible, real-time calcium monitoring, supporting preventive care, and guiding treatment plans for at-risk populations.

• Principal Investigator: Jiachen Zhang
• Website: https://sites.usc.edu/jzhang/
• Projects
  • Assessing Health Disparities Arising from Air Pollutant Emissions Linked to Port-Related Activities: A Case Study of the Ports of Los Angeles and Long Beach
    • Project Description: Port-related activities, encompassing the operation of vessels, trucks, and cargo handling equipment, are a significant source of air pollutants, leading to respiratory and cardiovascular diseases. Low-income, often minority-populated communities are more likely to be located near ports or freeways, and are therefore at a greater risk of exposure to air pollutants from port-related emissions. We propose a case study on the Ports of Los Angeles (LA) and Long Beach (LB), which have served as vital gateways for freight movement in the United States, handling 40 percent of the cargo entering the nation. This proposed research will develop a modeling framework to assess the air quality and health impacts of port-related emissions at a high spatial resolution and investigate the unequal burden of adverse environmental health impacts in disadvantaged communities. Utilizing this modeling framework, we will quantify the surge in primary particulate matter (PM) emissions resulting from port congestion in 2021, a situation driven by the COVID-19 pandemic and disruptions in the supply chain. Subsequently, we will for the first time assess the resultant health impacts of excess PM emissions on nearby communities. The insights gained from our research will provide insights for implementing measures aimed at mitigating the disproportionate exposure of disadvantaged communities to emissions from port-related activities.
    • Using machine learning approach to merge observational dataset of air pollutant concentrations
      • Project Description: This project aims to leverage machine learning techniques to effectively merge multiple observational datasets of air pollutant concentrations from various sources, including ground-based monitoring stations, satellite observations, and low-cost sensor networks. Air quality monitoring often relies on data from diverse sources, each with its own spatial, temporal, and methodological limitations. Integrating these datasets into a cohesive, accurate, and high-resolution dataset is crucial for better understanding air pollution patterns, exposure assessments, and supporting policy-making.
    • Open to: CURVE First-Time Researchers and CURVE Continuing Researchers
    • Student Responsibilities:
      • Collect and analyze weather and air quality observation data
      • Compare the model output with observational data and visualize the results
      • Collect and analyze massive amount of spatial data related to the emissions of locomotives and ocean-going vessels
      • Literature review in health impact assessment
    • Preferred Majors: Aerospace & Mechanical Engineering,Chemical Engineering,Civil & Environmental Engineering,Computer Science,Electrical & Computer Engineering
    • Preferred Skillsets:
      • Courses related to air pollution, climate change, and spatial data science
      • Programming skills (Python)
      • ArcGIS or QGIS
      • Abilities to handle big data

• Principal Investigator: Steve Nutt
• Website: https://composites.usc.edu/
• Projects
  • Facilities for materials characterization and testing, and for composite fabrication.

• Principal Investigator: Daniel McCurry
• Website: https://mccurrylab.com
• Projects
  • Contaminant removal and transformation in wastewater reuse systems
    • Project Description: Research activities in our lab focus on enabling safe water reuse by minimizing the presence of harmful contaminants in recycled wastewater. Our major research efforts include developing new chemical technologies for oxidation of trace organic contaminants, identifying the precursors and transformation pathways of disinfection byproducts formed during water treatment and wastewater reuse, and developing new analytical techniques for identification and quantification of chemical pollutants using mass spectrometry. Our research primarily takes place in the BHE Water Lab, and relies heavily chemical analytical instrumentation, including an ion mobility QTOF mass spectrometer, a gas chromatograph/triple quadrupole mass spectrometer, and an inductively-coupled plasma mass spectrometer. These tools allow us to quantify chemicals at extremely low concentrations (e.g., parts per trillion) and to identify previously unknown compounds in water samples.
    • Open to: CURVE First-Time Researchers,CURVE Continuing Researchers
    • Student Responsibilities: Assisting graduate students with experiments on water treatment, water reuse, and quantification of trace contaminants in water. Specific responsibilities will include preparing for and setting up batch water treatment experiments, sample preparation for mass spectrometry-based identification and quantification of pollutants, and analyzing data produced by analytical instruments. Especially dedicated students may eventually have the opportunity to advance to a fully-independent project advised directly by the PI.
    • Preferred Majors: Biomedical Engineering,Chemical Engineering,Civil & Environmental Engineering
    • Preferred Skillsets: General chemistry lab skills.
    • Some exposure to analytical chemistry is helpful but not required.
    • General chemistry coursework (e.g., CHEM105 here or AP equivalent)

• Environmental fate of wildfire-associated heavy metals
  • Project Description: Our recent research has found that aerial wildfire suppression potentially contributes toxic heavy metals to the environment. This project will involve bench-scale experiments and possibly field sampling to determine the fate and transport of these metals in the environment between soil and aqueous phases, and to evaluate the potential for surface water contamination by fire suppressant application.
  • Open to: CURVE First-Time Researchers,CURVE Continuing Researchers
  • Student Responsibilities: Assisting graduate students with experiments on the fate of heavy metals from wildfire suppressants applied aerially to wildfire sites. Specific responsibilities will include preparing for and setting up bench-scale soil column experiments, sample preparation for mass spectrometry-based and quantification of metals, and analyzing data produced by analytical instruments. The project may include field sampling of burn sites with evidence of aerial application of fire retardants, depending on fire season intensity and site access. Especially dedicated students may eventually have the opportunity to advance to a fully-independent project advised directly by the PI.
  • Preferred Majors: Biomedical Engineering,Chemical Engineering,Civil & Environmental Engineering
  • Preferred Skillsets: General chemistry lab skills
  • Some exposure to analytical chemistry is helpful but not required
  • General chemistry coursework (e.g., CHEM105 here or AP equivalent)

• Principal Investigator: Ruolin Li
• Website: https://ruolinli.me/
• Projects:
  • Re-engineering Transportation Systems for AV Integration
    • Project Description: The enhanced controllability of autonomous vehicles opens up a myriad of possibilities. For example, autonomous vehicles can be platooned with shorter headways, which increase road capacities, or they can act on information as altruistic decision-makers, thereby improving societal benefits. Autonomous vehicles and existing transportation infrastructure are also intricately linked. The challenge lies in managing or reshaping our current infrastructure to accommodate AVs efficiently and economically. How can we adapt our roads, signaling systems, and urban planning to meet the demands of AV technology? Insights into the dynamic interaction between robotic AVs and infrastructure optimization can guide us in creating cost-effective, efficient solutions that pave the way for the future of transportation.
    • Open to: CURVE First-Time Researchers ,CURVE Continuing Researchers
    • Student Responsibilities: Literature review, Coding numerical examples
    • Preferred Majors: Electrical & Computer Engineering, Industrial and Systems Engineering, Civil & Environmental Engineering, Computer Science, Aerospace & Mechanical Engineering
    • Preferred Skillsets: Interest in reading maths, Proficient programming for optimization in python or MATLAB

• • React and Interact: Harmonizing AVs with Humans
    • Project Description: Modeling human behavior in transportation systems is crucial yet challenging. Imagine drivers mischievously cutting in front of slowly moving autonomous vehicles just for fun! The impact of autonomous vehicles is significantly influenced by human reaction
    • Open to: CURVE First-Time Researchers ,CURVE Continuing Researchers
    • Student Responsibilities: Literature review, Running SUMO simulations
    • Preferred Majors: Civil & Environmental Engineering, Electrical & Computer Engineering, Industrial and Systems Engineering, Computer Science, Aerospace & Mechanical Engineering
    • Preferred Skillsets: Interest in reading maths, Basic programming experience in python, Prior experience with traffic simulation softwares, especially SUMO

• Principal Investigator: Adam Smith
• Website: https://www.smithresearchusc.com
• Projects
  • Investigating Bioelectrochemical Systems as Biosensors in Engineered Water Treatment
    • Project Description: The project uses exoelectrogenic microorganisms to generate an electrical signal in the presence of organic matter. This can be utilized as a biosensor to provide real-time monitoring of aggregate organics in engineered systems, such as bioreactors used for wastewater treatment.
    • Open to: CURVE First-Time Researchers
    • Student Responsibilities: The student will assist a PhD student in operating and monitoring bench-scale bioelectrochemical systems. The student will be able to learn analytical chemistry and molecular biology techniques.
    • Preferred Majors: Chemical Engineering,Civil & Environmental Engineering
    • Preferred Skillsets: Chemistry and biology
  • Investigating per- and polyfluoroalkyl substances (PFAS; forever chemicals) in various environmental matrices and bench-scale bioreactors
    • Project Description: This project investigates the fate of PFAS across water reclamation processes at the full-scale and in bench-scale systems in the lab. We also quantify PFAS in a community engagement process focusing on drinking water and dust samples.
    • Open to: CURVE First-Time Researchers
    • Student Responsibilities: The student will assist PhD students in operating and monitoring bench-scale bioreactors and quantifying PFAS using analytical chemistry.
    • Preferred Majors: Chemical Engineering,Civil & Environmental Engineering
    • Preferred Skillsets: Chemistry and biology

• • Using Membrane Biofilm Reactors to Discover Novel Microorganisms that Transform Greenhouse Gases
    • Project Description: Membrane Biofilm Reactors allow for delivery of a gas through a membrane with a biofilm adhered to the surface that can utilize that substrate. We are using these systems to enrich and isolate for novel microorganisms that transform greenhouse gases inclu
    • Open to: CURVE First-Time Researchers
    • Student Responsibilities: The student will assist with bioreactor operation and monitoring using analytical chemistry and molecular biology techniques.
    • Preferred Majors: Civil & Environmental Engineering
    • Preferred Skillsets: Chemistry and biology

• Principal Investigator: Jesse Thomason
• Website: https://glamor-usc.github.io/
• Projects
  • GLAMOR: Grounding Language in Actions, Multimodal Observations and Robots
    • Project Description: We bring together natural language processing and robotics to connect language to the world (RoboNLP). Our lab is broadly interested in connecting language to agent perception and action, and lifelong learning through interaction.
    • Open to: CURVE First-Time Researchers
    • Student Responsibilities: Attending weekly research planning syncs with a PhD student mentor, occasionally presenting in the GLAMOR lab meeting about progress, writing research software, designing and conducting experiments.
    • Preferred Majors: Computer Science
    • Preferred Skillsets: Fluency with python, interest in or knowledge of deep learning and machine learning coding libraries, interest in or knowledge of robot hardware and software systems like ROS.

• Principal Investigator: Heather Culbertson
• Website: https://sites.usc.edu/culbertson/
• Projects
  • Haptics for Virtual Reality and Social Interactions
    • Project Description: Haptics Robotics and Virtual Interaction (HARVI) Lab investigates touch based interactions between humans and technology. The student will be immersed in an exploration of the Human-computer Interaction (HCI) based research, with a special emphasis on Virtual Reality (VR) and social haptics. The project will facilitate the acquisition of knowledge about the fundamentals of VR, software, and algorithms that enable digital real-time interactive communication. It will also give insights into wearable systems aimed to create touch illusions.
    • Open to: CURVE First-Time Researchers
    • Student Responsibilities: Student is expected to gain hands-on coding experience, display a keen interest in developing the technical skills (Unity 3D, C#, App development) required to design, implement, and troubleshoot game engines and VR platforms.
    • Preferred Majors: Aerospace & Mechanical Engineering,Computer Science,Electrical & Computer Engineering
    • Preferred Skillsets: Prior experience with algorithms in Unity or any other game engine is preferred, but not a requirement. The student's motivation, comprehensive approach to learning, combined with a deep curiosity for cutting-edge VR technology and wearables, forms the foundation for future academic pursuit in Human-computer Interaction.

• Principal Investigator: Emilio Ferrara
• Website: https://election-integrity.online/
• Projects
  • 2024 Election Integrity Initiative
    • Project Description: This project aims to leverage publicly available datasets from social media and political ads to analyze patterns and discover insights into electoral processes. By using data analysis and machine learning practices, the project seeks to protect democratic processes by monitoring and analyzing online media to identify and combat malicious interference. We strive to ensure the integrity of elections and public discourse through cutting-edge research and advanced echnological solutions.
    • Open to: CURVE First-Time Researchers,CURVE Continuing Researchers
    • Student Responsibilities:
      • Data Collection and Preprocessing of Databases
      • Data Analysis: Conduct exploratory data analysis to identify trends, patterns, and anomalies within the datasets.
      • Machine Learning Model Development: Develop and train machine learning models to detect irregularities in electoral data.
      • Data Visualization: Create interactive visualizations to present findings in a clear and accessible manner.
      • Report Writing: Document methodologies, findings, and insights in a comprehensive report.
      • Collaboration: Work collaboratively with team members and mentors, participating in regular meetings and updates.
    • Preferred Majors: Computer Science
    • Preferred Skillsets:
      • Data Science: Proficiency in data analysis and machine learning, with coursework or experience in statistics, data mining, and predictive modeling.
      • Programming Skills: Experience with programming languages such as Python or R, particularly in data manipulation and machine learning libraries (e.g., Pandas, Scikit-learn, TensorFlow).
      • Data Visualization: Knowledge of data visualization tools and libraries such as Tableau, Matplotlib, or Plotly.
      • Critical Thinking: Ability to think critically about data and methodologies, ensuring the integrity and validity of findings.
      • Communication Skills: Strong written and verbal communication skills to document and present findings effectively.

• Principal Investigator: Maja Mataric
• Website: https://uscinteractionlab.web.app/
•  Projects
  • A Haptic Feedback Device for Guided Mindfulness Exercises with a Socially Assistive Robot
    • Project Description: Prior studies have explored the applications of both physical feedback devices and socially assistive robots (SARs) to aid users during mindfulness exercises, but no study has explored the potential of combining these two modalities during mindfulness practice. This work aims to design a physical device that gives appropriate haptic feedback through motor movements to guide users through a mindfulness exercise led by a socially assistive robot. Students working on this project will generate and prototype designs for the device that meet the following goals: low-cost and easily fabricated by non-expert users, cohesive with the design of the Blossom robot platform, and easily synced to different mindfulness exercises.
    • Open to: CURVE First-Time Researchers,CURVE Continuing Researchers
    • Student Responsibilities: Students will use an arduino microcontroller and various electronic components to build and program the device. They will explore existing haptic mindfulness devices and design a device for the Blossom robot. They may use different fabrication techniques to build the device such as 3d printing, laser cutting, or crochet.
    • Preferred Majors: Aerospace & Mechanical Engineering
    • Preferred Skillsets: CAD modeling (especially Fusion 360), 3d printing, technical writing

• Designing Compelling Robotic Study Companions for Students with ADHD
  • Project Description: Prior studies have shown that college students with ADHD respond positively to simple social robots that monitor attention and provide immediate feedback during periods of inattention or impulsivity. In feedback from a recent pilot study of in-dorm study companion robots, participants identified increased interactivity and positive rewards as potential ways to motivate users to complete more study sessions with a study companion robot. This work aims to incorporate different modes of interaction (e.g. touch, speech) and positive rewards (e.g. visual light displays, opportunities to customize the robot) on a simple, low-cost social robot in an effort to increase adherence to regular study sessions. We will evaluate these behaviors by measuring their effect on students' engagement as they complete a schoolwork task.
  • Open to: CURVE First-Time Researchers
  • Student Responsibilities: Students will use python to program behaviors for a Blossom socially assistive robot. They will also use electronic components such as touch sensors and LED lights to add functionality to the robot. They will design a user study to evaluate these added capabilities and measure their effect on user adherence.
  • Preferred Majors: Aerospace & Mechanical Engineering
  • Preferred Skillsets: CAD modeling (especially Fusion 360), python.

• Peer Mediation Practice with Socially Assistive Robots for Elementary School Students
  • Project Description: Peer mediation is a model of conflict resolution used in many K-12 schools across the United States. This model lets students serve as peer mediators to help other students resolve conflicts. Peer mediation has been shown to be an empowering form of conflict resolution for the mediators; however, mediators are usually chosen by teachers and peers for their leadership quality and positive social behavior. For students who are not usually chosen to be peer mediators, or for those who do not have access to a peer mediation program, socially assistive robots (SARs) may provide a way for practicing this valuable skill. The goal of the study is to help evaluate the use of SARs in peer mediation role-play practice in improving self-perception of 3rd to 5th grade students. The study insights will help determine if SARs are a useful tool in helping 3rd to 5th graders practice conflict resolution skills. The collected data will help determine how to improve SARs to make the interaction more realistic and impactful.
  • Open to: CURVE First-Time Researchers
  • Student Responsibilities: Students will assist with data collection on-site at participating elementary school (this will involve setting up camera equipment and setting up Raspberry Pi computers connecting with the Blossom robot) curating and maintaining video, audio, and structured data, statistical analysis of data, reporting findings in an academic paper.
• Preferred Majors: Computer Science, Electrical & Computer Engineering
• Preferred Skillsets: Python programming; experience with Amazon Web Services (AWS) applications for storage (S3) and full stack app development; experience with data analysis; experience with statistical analysis; experience with Raspberry Pi and internet of things (IoT); any experience with social/behavioral projects is a plus.

• Multimodal Models of Engagement and Cohesion in Group Therapy
  • Project Description: Group interaction is a many-step process with multiple interactors. An autonomous facilitator requires an understanding of the processes and interaction states of the group. We propose to develop automated models for understanding individual and group processes at three levels, as follows.
    1. Speaker. An autonomous facilitator should be able to identify the active speaker(s) and facilitate turn-taking in the group.
    2. Interaction and engagement. Automated understanding of social processes in group interaction is an important building block for an autonomous facilitator. A group facilitator should understand the context and state of the conversation to guide it effectively. Moreover, individual-level measurements of engagement are important for recognizing higher-level group constructs, like group cohesion. In the context of this proposal, this will involve automated coding of interaction.
    3. Group cohesion. Group cohesion is defined by the level of social connectedness in a group. An automated measurement of cohesion will require understanding emotional and social engagement and group structure.
  • Open to: CURVE First-Time Researchers, CURVE Continuing Researchers
  • Student Responsibilities: Develop multimodal models (in Pytorch) of engagement and group cohesion from a dataset of group therapy interactions with a robot facilitator. Clean and segment study data. Write academic papers with findings.
  • Preferred Majors: Computer Science
  • Preferred Skillsets: Python, data processing, technical writing

• Socially Assistive Robot-Administered Cognitive Behavioral Therapy
  • Project Description: There is a mental health epidemic that is disproportionately impacting university-age young adults. However, mental health systems are strained and unable to meet the demand for access to care. There are also a multitude of barriers to receiving mental health care, such as accessibility and stigma. Thus it is important to develop technologies that can assist in overcoming these barriers. This project aims to develop a Socially Assistive Robot (SAR) that can guide students through Cognitive Behavioral Therapy (CBT) at-home exercises. We will develop SAR-guided CBT exercises through co-design sessions with expert therapists and test them through a user study with university students.
  • Open to: CURVE First-Time Researchers, CURVE Continuing Researchers
  • Student Responsibilities: Develop movements and behaviors on the Blossom Robot. Language model prompt engineering. Develop SAR-guided CBT exercises in collaboration with trained therapists. Learn to run co-design sessions as well as a user study to test the exercises.
  • Preferred Majors: Computer Science
  • Preferred Skillsets: Python programming, AWS, ability to work directly with participants

• Using Multimodal Machine Learning to Model Early Life Stress in Infancy and Early Childhood
  • Project Description: Adverse Childhood Experiences (ACEs) can include violence, abuse, and growing up in a family with mental health or substance use problems. ACEs are linked to chronic health problems, mental illness, and substance misuse in adulthood. Using data collected by the Children’s Hospital of Los Angeles’ (CHLA) Levitt Lab, we will model ACEs using audio and visual interactions between mothers and children, mother emotional state questionnaires, mitochondrial data, heart rate variation, and other modalities. Additionally, students may be involved in the creation of socially assistive robots to add social and emotional learning interventions in early childhood to mitigate the social and emotional negative effects of ACEs.
  • Open to: CURVE First-Time Researchers, CURVE Continuing Researchers
  • Student Responsibilities: Students will use Python to collect, clean, and model multimodal information collected from CHLA. This effort will include the use of classical machine learning, as well as more advanced multimodal fusion modeling techniques. Students might also use the QTRobot to develop interventions and collect affective data such as audio and video for further multimodal analysis.
  • Preferred Majors: Computer Science, Electrical & Computer Engineering
  • Preferred Skillsets: Experience with python and some experience with data or statistics. Preferred – not required.

• Play and Learn: Design Socially Assistive AI for Children with Speech Articulation Needs and Their Care Ecosystem.
  • Project Description: It’s estimated that almost 50%, around 6 million, of U.S. children with speech disorders went untreated in the past year, negatively influencing their academic performance and social interactions [1]. This highlights the urgent need for more accessible speech therapy options. For children undergoing speech therapy, adherence to low-dose high-frequency in-home practices is vital for treatment efficacy, yet many parents, already stretched thin by their responsibilities, find it challenging to provide consistent engagement. for their children. In this project, we aim to simplify speech therapy practices for families with personalized learning powered by socially assistive AI.
  • Open to: CURVE First-Time Researchers, CURVE Continuing Researchers
  • Student Responsibilities: Developing the software agents and robots either in React Native or in Python; Helping deploy and evaluate the software agents and robots; Data collection for a speech dataset in the setting of speech articulation learning; Training and evaluating machine learning models for phoneme recognition models.
  • Preferred Majors: Computer Science
  • Preferred Skillsets: React Native, Python, Pytorch

• Principal Investigator: Erdem Biyik
• Website: https://liralab.usc.edu/
• Projects
  • Robot Learning from Natural Human Feedback
    • Project Description: Most of the existing robot learning algorithms rely on human feedback that is explicitly designed to teach the robots. However, humans leak a lot of information even when they are not trying to teach. For example, their gestures, gaze, facial expressions etc. carry information about what they want the robot to do. The goal of this project is to enable robots use such natural feedback by utilizing machine learning techniques.
    • Open to: CURVE
    • Student Responsibilities: Implementation of algorithms, deployment on simulation and real robot platforms.
    • Preferred Majors: Computer Science
    • Preferred Skillsets: machine learning, Python

• Principal Investigator: Murali Annavaram
• Website: https://scip-lab.usc.edu
• Project #1
  • Project Title: Resource Efficient Algorithm and System Design of Large Language

Models:

• • • Project Description: Large Language Models (LLMs), including GPT-4, LLaMA, and
    PaLM, have demonstrated exceptional capabilities in natural language
    understanding, generation, and complex reasoning tasks. Despite their impressive
    performance, these models face significant challenges due to their substantial
    computational and memory demands during training and inference. Addressing
    these efficiency challenges is crucial for the broader adoption and application of
    LLMs. This project focuses on making LLMs more efficient in training and inference
    by reducing their high computational and memory demands. The goal is to develop
    and test new algorithms and system-level techniques that optimize resource use,
    enabling broader and more sustainable applications. Depending on interest,
    students may work on algorithmic improvements, system optimizations, or model
    compression techniques.
    • Student Responsibilities: Investigate the state of the art, reproduce experiments,
    and improve and implement ideas.
    • Interview Required? Yes
    • Preferred Skillsets:
    o Python
    o Pytorch/Huggingface
    o Analytical Skills
    o Machine Learning
    o Natural Language Processing
    o LLM Architectures.
    • PhD Mentor: Hossein Entezari Zarch <entezari@usc.edu>

Project #2

•  Project Title: Energy Implications of Training and Serving Deep Learning Models
  # of Freshmen Positions: 1
• # of Continuing Student Positions: 0
  •  Project Description: This is an ongoing research project where the student will take
    part in measuring and analyzing the energy consumed by deep learning training and
    serving pipelines. As there is no agreement in the literature on how to measure the
    energy precisely while using Machine Learning backends like PyTorch, we plan to
    conduct multiple experiments to build a stronger case for our proposed
    methodology. Our primary focus will be on recommendation models and large
    language models, which could be extended to other classes of deep learning
    models. This project could result in a publication that will pave the way for the
    student to pursue a research career.
    •  Student Responsibilities: Measure the energy budget for training and serving various
      deep learning models
      • Interview Required? No
      • Skills and Competencies:
      • Knowledge of UNIX environment
      • Programming Experience in Python and C++
      • Decent Experience in PyTorch
      • Good analytical skills
      • Applied knowledge of recent Deep Learning Models (LLMs, Rec Sys)
      • PhD Mentor: Yongqin Wang <yongqin@usc.edu>