Partner: AWS

Ulises Maldonado

Xiling Tian

Neeraja Beesetti

Jessica Bhalerao

David Horta

James Kwan

Project Overview: There are many competitors to Amazon Connect. However, most of them share the same pitfall when it comes to supervisors. The dashboards are overly complicated and designed for administrators instead of supervisors. They choose to go down the route of presenting you with all the information all at once. This has the downside of cluttering the web apps and making it hard to navigate. Our implementation will take note of these issues and trim Amazon Connect. We do not seek to replace Amazon Connect. We do not seek to expand on Amazon Connect. Instead, we plan to take the relevant parts of Amazon Connect (the data that matters to supervisors) and combine it with custom metrics we create (and a machine learning model if time allows) to provide a dedicated supervisor management app for Amazon Connect instances.

Partner: CHOC

Kenny Yu

Marcus Linture

Calvin Nguyen

Kelly Hu

Evan Leong

Project Overview: Currently, there is research to recognize hemodynamic instability in patients using machine learning. A Hemodynamic Stability Index (HSI) was developed to determine risk for potential hemodynamic interventions. While this research is closely related to our problem, this was aimed at providing better bedside care in the intensive care unit for all hemodynamic instability issues. While that research was for more broad care, our focus in this project is dedicated to a specific injury to childrens’ solid organs. The traditional way of determining hemodynamic stability for these injuries is dependent on the doctors’ expertise and quick judgment, but our project will help give a faster prognosis to make life-saving decisions.

Partner: Gotchoo

Jiayun Wang

Hang Cao

Nhat Anh Pham

Audrey Lam

Carr Jeffrey Pidlaoan

Project Overview: Businesses tend to have their own solution to give people options to gift their products to others by using their own technologies. Some users also have to rely on third-party technologies to request and send gifts to and from others. Gotchoo plans to give a centralized solution to users. Businesses would add their “gift it” button to their websites and Gotchoo will handle the backend and fulfillment to ensure the products users flag ends up in a centralized database and viewing system.

Partner: NASA ASU

Jingjing Wang

Tracy Trinh

Mendel Shifrin

Carl Grini

Jayden Le

Trey Miller

Project Overview: Our project is to create an immersive, 3D, web experience that can be accessed through scanning a QR code. The experience will serve as a tool to educate individuals of all ages about the NASA Psyche mission. Psyche is an asteroid that orbits the Sun between Mars and Jupiter. Unlike all other asteroids mankind has studied, Psyche is made mostly of metal rather than rock or ice. Because of this, NASA believes that it might be the early stage of a planet’s core.

We can’t access or study Earth’s core with today’s technology, so studying Psyche might provide answers on how Earth’s core and other cores came into existence. Our immersive experience will help further the knowledge and highlight the importance of the Psyche mission to the public. By providing an interactive visual experience, it will help people of all ages understand the Psyche mission better. It might leave lasting impressions that may inspire future generations down the line to work in a similar field. It also will highlight the importance of funding programs like NASA, and act as transparency to the public of how the money given to them is being spent.

Partner: Paciolan Inc

Levi Varga

Dylan Vu

Christopher Kwong

Casey Tran

Duong Vu

Henry Reyes

Project Overview: Paciolan handles sensitive user payment information, as they are a ticketing ecommerce platform and service. In order to maintain payment security, Paciolan regularly generates new “secret keys” that are used to process user payments through a payment processor. This secret key is kept in the company’s database, and needs to be manually verified and copied over to the database. The current process of doing this is highly manual and error prone. If an error occurs while updating the keys, it causes downtime, leading to lost profits and a worse user/customer experience.

Our project aims to automate this secret key verification and updating the workflow to reduce the possibility of human error during the updating of the payment secret keys for Paciolan through a solution that is fully integrated and compatible with the company’s current tech stack and codebase.

Partner: Partner ESI

Brian Le

Akyra Lee

Lucas Murray

Jason Yang

Jingxuan Dou

Project Overview: There are many open-source PDF parsers that can read text from PDFs. These parsers are available in various languages such as Python, Java, and Javascript. There are also more advanced AI PDF parsers that utilize optical character recognition (OCR) to read from images and unformatted text and translate characters into different languages. Some commercial examples include Amazon Textract, ChatGPT, or Apryse. However, with our project, our AI PDF parser parses data given a PDF report containing text, tables, and sometimes images and organizes it into a vector data point inside a vector database. There is no tool that can specifically parse specific data Partner wants to store inside their database. Additionally, no AI PDF parser can parse correctly given the different formatted PDF reports PartnerESI has been using over the past 10 years.

Partner: Partner ESI

David Lim

Touch Visa

Mitsutoshi Sato

Natali Perrochon

Laura Valko

Project Overview: When buyers consider purchasing a commercial property, it is important for them to be informed about a wide variety of factors to do with the property. Historical records and images, along with visits to the property and soil tests, can be compiled in a report that the customer can use to make a decision regarding their purchase. However, as time passes, these properties can undergo changes which need to be reflected in future reports. Looking over every existing source of information each time a property is on the market is inefficient, especially if this same research was done before and only needs minor updates. Partner Engineering and Sciences has implemented an online system that converts old reports to new reports by selecting modifications to the property, allowing minor changes to be done easily and efficiently. However, this solution is limited to reports created within the system, meaning that all of the old reports that may be used still have to be manually reviewed. This process of inputting the data from the report into the system is tedious and slow. This means this data is often excluded from analytics. To speed up this process and allow for a larger online database, our project aims to create a machine learning algorithm to extract important field information from old reports and format them in a JSON file that can be downloaded and easily uploaded into their online system.

Partner: Partner ESI

Andy Quan Nguyen

Dustin Khang Leduc

Kevin Pham

Nero Li

Thien Toan Vu

Project Overview: Currently, Partner ESI has mass amounts of data and visualizations powered by PowerBI. They also have a lot of unstructured data that is not a part of that. Using their data, Partner ESI wants to create an accurate, interactive chatbot for clients to query for different types of information, which will ultimately give the clients a
much better experience with the product. Some examples include clients being able to get maintenance and estimated prices/times for maintenance for an individual property or a portfolio, clients or employees being able to submit documents and the model summarizing it and getting the main points of it, and looking at property characteristics. This will also be time efficient for the company and clients because they would not have to go through lengthy documents. This project will make life more convenient and efficient for both the employees and clients.

Partner: SAP ERP

Yue Wu

Robert Lauv

Angeline Hui

Jacob Wong

Senghoung Lim

Project Overview: Currently there are frameworks such as DORA and SPACE to improve efficiency in sprints, want a more data driven approach, more focused on the perspective of the end users (scrum teams developers and product managers) than stakeholders, more tailored to the scrum team.

Also, the current team that we will meet is called the Honey Bear team and they are currently in development of metrics such as revenue growth optimization which inputs various variable which affect revenue and helps manage these variables to boost long-term revenue growth

Kaifan Yu

Jonathan Tran

Michelle Luu

Taylor Quach

Project Overview: Currently, professors have to read through 50+ pages of student evaluations given to them in a pdf or csv file. The sheer magnitude of responses makes it difficult for professors to go through all student’s input. The student responses are also ordered by longest to shortest which tends to result in upset rants at the top. As a result, professors are resorting to trading evaluations with other professors to get a summarization of the student’s sentiments. These problems dissuade professors from actually using evaluations as a tool to improve their class.

Several different organizations and technologies are available that try to minimize these problems. The EEE suggests that professors contact the DTEI (Division of Teaching Excellence and Inclusion) for analysis and advice regarding the responses from the evaluations. However, the DTEI focuses more on emphasizing diversity and demographics than specific course feedback.

Additionally, the project, PERI^2, exists which allows professors to pay for a research and data analysis consultant that will provide an in-depth examination of feedback. There are also current technologies, like ChatGPT, where a document can be fed in to extract summaries and key words. However, it stores all your inputted data which is a concern for the protected student evaluation responses.

Our advances would be automating the feedback analysis process to reduce or eliminate the need for manual parsing of surveys, expanding this capability to process multiple documents, and catering to professors with diverse backgrounds.

Partner: TCS

Ngoc Minh Chau Ho

Andre Saldanha

Huu Le

Andrew Fleischer

Jayson Nguyen

Project Overview: The main issue that we are resolving is data privacy and protection. Many companies are reluctant to share data with other parties due to fear of having their data breached. This affects many different areas such as AI model building, which requires data to be inputted to train it.

One of the solutions to this problem is confidential computing, which isolates and protects data while it is being processed. Numerous virtual machines support confidential computing, such as Google Cloud and AWS Nitro Enclaves. Our project will remedy this issue by extending the life cycle in which the data is being protected. This will include encrypting the data during processing and then storing it within an enclave so that it can only be viewed by approved parties.

Partner: TCS

Alex Figura

Hai Nguyen

Mei Yifan

Amy Huang

Myron Zhou

Project Overview: The current landscape for synthetic data generation primarily relies on domain-specific algorithms and generative adversarial networks (GANs) or large language models to produce data, namely formatted text and tabular data, that mimics real-world distributions. While effective in certain contexts, these solutions often face challenges in generating complex, multi-modal data (image, database, etc) and ensuring minimal data hallucination, which will create non-realistic or implausible data points that compromise the data quality.

Our project introduces a reinvention by using a multi-agent network large language model (MA-LLM) framework designed to address these limitations. This approach enables the generation of high-quality, cross-modality synthetic data — including images and direct formation of the database as bulk— with greater adherence to real-world distributions and contexts while providing maximized efficiency. By integrating multi-agent specialized in different data modalities, our platform significantly reduces the occurrence of hallucinations, ensuring that the generated synthetic data is both realistic and diverse.

Partner: Zuum Transportation

Priyansha Sharma

Ryan Tighiouart

Howard You

Sahil Randhawa

Meenakshi Mynampati

Project Overview: Zuum provides a one-platform solution for shippers, brokers, carriers, and drivers to manage orders and finances, receive actionable intelligence, and track in-transit trucks. On this platform, we want to offer brokers an effective and intuitive way to retrieve information about their clients and truckers. Through a NLP chatbot, brokers will no longer have to learn complex interfaces to get logistics information; instead, they will be able to easily fetch real-time information about in-transit trucks from the backend with natural language. Providing this information to brokers enables them to save time, make smarter supply-chain decisions, and save money on logistics. Chatbot driven information retrieval also allows a business to increase overall productivity and efficiency, handling multiple requests, providing 24/7 availability, performing repetitive tasks, and minimizing human errors such as typos and miscommunication.

Partner: Raytheon Technologies Corporation – RTX

Diego Cortez-Tena

Seraphine Wu

Seunghyun Choi

Raymond Chou

Kalman Wong

Project Overview: Validating the legitimacy of products is an increasingly important job in today’s day and age where things can be manipulated easily and cyber attacks are getting more frequent and fierce. RTX currently has a product verification process called Cyber Verification Lab (CVL) that is still at a prototypical stage. We want to create a system that integrates and supports this testing process. It would provide a user interface that would help streamline the workflow of the security team and help aggregate information from multiple testing processes and roles in the security team. The system will help make the testing process more repeatable and applicable for products in various forms such as software, hardware, images, textual/graphical reports, etc.