iSchool Capstone

2025

Project Logo

AI Powered Data Quality Checker

Our project, sponsored by Included AI and Info Edge, focuses on enhancing data reliability through an AI-powered data quality checker. The solution identifies anomalies, standardizes inconsistent entries, and visualizes data integrity metrics, ensuring clean and trustworthy datasets for business insights. We developed automated pipelines that detect and correct real-world data issues, such as inconsistent location naming and missing fields. An intuitive dashboard allows stakeholders to monitor data quality over time. By integrating AI with usability, our tool reduces manual cleaning effort, increases trust in analytics, and empowers data-driven decision-making across enterprise teams.
Project Logo

MatchMind – Intelligent hiring tool powered by LLMs, RAG and GraphDB

We leveraged the nascent AI native RAG and Graph Database to develop a resume screening and shortlisting tool. Traditional ATS systems often present challenges for hiring managers and recruiters. Strict text-matching methods may screen out capable, qualified candidates who word their resumes differently from the job description. To address this issue, we employed a GraphDB to uncover latent relationships among entities, enabling a more semantic understanding of candidate profiles. Additionally, we packaged these capabilities into a chatbot interface, allowing hiring managers to interact with resumes and job descriptions using natural language.

2024

Project Logo

Pixel Pioneers: Expanding a nonprofit’s AI medical screening abilities to Tuberculosis

Low income populations often lack access to health insurance that enables them to test for illnesses. Tuberculosis alone carries an estimated price tag of $287 per test in the United States. In an effort to counter this out-of-pocket expense, we constructed an advanced machine learning model that predicts the likelihood an individual has Tuberculosis with high accuracy. The prediction is based on symptoms, demographics, and a self recorded cough recording, and the process is quick and easy with the help of a chatbot. By using our application, individuals can make an informed decision about whether medical intervention is necessary.
Project Logo

A Gentle Bibliographic Madness: Cataloging and Digitizing the Journal of the Book Club of Washington

Working for the Book Club of Washington, a bibliophile society dedicated to book collecting and book arts, I created an online public access catalog for the organization's independently published periodical, the Journal of the Book Club of Washington. To do so, I used LibraryThing to create bibliographic records for over 300 of the Journal's articles from its debut in 2000 to Spring 2023. I then organized those records into an OPAC utilizing LibraryThing's TinyCat software. Additionally, I digitized each issue that lacked a digital copy and made the digital versions available to the public via the online catalog.
Project Logo

A Web for All: Generative AI Powered Navigation for the Neurodiverse

Nav-Easy.AI addresses the widespread lack of web accessibility for neurodiverse users, with 96% of websites failing in full inclusivity. Unlike existing tools that target specific neurodiverse conditions, Nav-Easy.AI offers a universal, AI-driven solution that caters to the broad spectrum of neurodiversity without needing to modify the web user interface. It facilitates voice or text automated web interactions and navigation, reducing cognitive overload for all. Just say or type what you need, and Nav-Easy does it for you. With Nav-Easy.AI, inclusivity is more than a buzzword, it's a reality.
Project Logo

A Window To Seattle’s Music History: Improving Access to the Ladies Musical Club’s Archives

The Ladies Musical Club (LMC) is Seattle’s oldest arts organization. Founded in 1891, LMC helped establish Seattle as a musical city by giving skilled women musicians a platform for public performance, thrilling audiences with concerts by world-famous musicians, and supporting other Seattle institutions like Cornish College. LMC’s important history is well preserved but can be hard to find. This project aims to improve accessibility to the LMC’s archive by providing a cross-institutional collection guide and enhancing the LMC’s digital archive. These changes offer a clearer window to Seattle’s music history for LMC members, researchers, music fans, and the broader public.
Project Logo

Accessibility in Reader Services: Obtaining a Reader Card at The Huntington Library

In July of 2021 the Reader Services department at The Huntington Library expanded their access policy, resulting in a more diverse population of patrons. This project aimed to determine what services could be improved or employed to better serve Huntington researchers through an accessibility assessment of the special collections reading room and the online process users must undergo to obtain a reader card. The assessments revealed a need to create intentional space for readers with disabilities, both in person and online. Recommendations included reading room alterations, the implementation of an accommodations request libguide, and updated language throughout the library's website.
Project Logo

Accessing Memory: Preparing a Digital Archive for Vanport Mosaic

Vanport Mosaic is a community archive that preserves memories related to life in Vanport, Oregon, the flood that destroyed the city, and the reverberating legacies of displacement, discrimination, resistance, and resilience that followed. This project addresses Vanport Mosaic’s need to make its content available online by developing a rights and usage policy, designing a metadata schema, and populating descriptive metadata for the collection’s oral history interviews to increase discoverability. Through this work, the marginalized histories in this collection will become discoverable and accessible to the public, making them available to educate, inspire, and advocate for change.
Project Logo

AI Potential with Service Requests and Incidents

Within the City of Seattle, the IT Department resolves service requests and incidents through service/help desk or the Service Hub system. Given the complexity of the data generated from the system, it's challenging for IT staff to extract actionable insights to expedite the resolution of stalled requests. This project aims to transform organizational management embedded with AI techniques. By implementing Llama 2 with the LangChain framework and running open-source large language models locally, we developed a chatbot (EVAN9000) to optimize operational efficiency while ensuring data security. Extensive experiments and evaluations have validated its effectiveness for promising enterprise-level use cases.
Project Logo

AI4AI

Our project addresses the challenge of distinguishing authentic news from misleading content in today's rapidly evolving digital landscape. Utilizing advanced machine learning models—SVM, Logistic Regression, Naive Bayes, and XGBoost—we compared their efficacy in accurately classifying news articles into four categories: Human Real, Human Fake, Machine Real, and Machine Fake. Our results highlight SVM's accuracy and stability, making it the most effective model for this task. By enhancing the detection and prevention of misinformation, our project significantly contributes to public access to credible news, ensuring that individuals can make informed decisions in a world increasingly influenced by AI-generated content. This effort not only enhances public information integrity, but also addresses the essential need for instruments that can prevent the sophisticated spread of false news.