Roger Nguyen

Learn more Tech profile

About

Roger Nguyen is a rising honors student in Computer Science at Villanova University. He is also exploring a double major in Economics and a concentration in Cognitive Science, underlining his enthusiasm for the intersection of technology, economics, and human cognition.

His deep-rooted interest in Education propels his vision: applying Cognitive AI in behavioral economics and personalized STEM Education. His long-term goal is to develop an ecosystem where global children can access personalized learning with high-quality and zero-cost science resources.

Yours sincerely,

Roger Nguyen

Experiences

NLP Research Assistant

Villanova Center for Research and Fellowships

May 2024 - Current

• Leveraged BERTopic, a topic modeling technique, to categorize medical interviews, utilized LLMs to generate meaningful labels from keywords, and executed tasks on Augie (Villanova’s HPC cluster) with Linux commands.

• Finetuned small-scale language models by filtering 10M words from 100M English-German translations, based on X-Comet quality estimation and curriculum learning, improving BLiMP and EWoK scores by 0.5-6 points.

• Co-authored a paper submitted to 2024 BabyLM Challenge, with implications for low-resource languages.

Software Engineer Fellow

HeadStarter AI

Jul. 2024 - Sep. 2024

• Led a 3-fellow team to develop 5 AI-powered SaaS products using TypeScript, Next.js, Firebase, and Vercel, following Agile methodologies with weekly sprints and incorporating CI/CD practices for iterative deployment.

• Optimized a web scraper for Rate My Professors and upserted to Pinecone index integrated with RAG pipeline using LangChain and Gemini 1.5 model to get up-to-date answers, reducing query response times by 25%.

• Contributed 10 pull requests to build microservices and increase unit tests coverage by 20% using Playwright.

VR/Data Research Assistant

Human-Centered Cities Lab

Jan. 2024 - Current

• Designed external human-machine interaction systems for autonomous cars to enhance pedestrian communication.

• Leveraged large language models to analyze situational context, enhancing decision-making accuracy by 25%.

• Analyzed driver-vehicle psycho-behavioral data using Matplotlib and Pandas, saving researchers 20% time in data interpretation.

Software Engineer Intern

FPT Smart Cloud

Jun. 2024 - Aug. 2024

• Built a performance dashboard for 200 employees using React, Flask, Chart.js, REST API, and MVC design.

• Trained in SDLC like containerization with Docker or CI/CD with GitLab, reducing deployment time by 30%.

• Boosted decision-making and cloud app development by 20% in 25-member data fusion team with Agile.

Projects & Papers

Curious learner of new programming languages and frameworks for side projects.

Passionate programmer most familiar with: Python, HTML, CSS, and JavaScript.

Vintage Tennis

• Developed a tennis-themed social media platform using React, HTML/CSS, and JavaScript, enabling post creation, content searching, and sorting with UUID-based user authentication.

• Enhanced interactivity with local image uploads and real-time updates via Supabase, streamlining user experience with loading animations for seamless data fetching.

VIEW CODE

Multi-Dimensional Sentiment Analysis

• Collaborated with 2 AI4ALL Ignite Fellows to utilize diverse datasets to improve interpretative AI capabilities for applications in social media analytics and linguistic research.

• Developed an emotion predictive model, implementing a Random Forest Classifier combining features of raw sentiment score, text vectorization, scaled text length, and word count, achieving an accuracy of 0.8605 and an F1-score of 0.8587.

VIEW CODE

Positive vs. Negative

• Developed a website to analyze attitudes using Flask for dynamic pages and algorithms for efficient content delivery.

• Implemented Transformers model for sentiment analysis, hashing for users' privacy, and SQLite for storing users' portfolios.

VIEW CODE

Vietnamese Fake News Detection

• Selected as top 15/200+ applicants for 2-week intensive ML/DL learning & operated AI-related games in MaSSP Open Day.

• Discovered tokenization process from Bartpho model, collaborated on training dataset of ReINTEL containing 10,000+ labeled examples, & achieved promising results with 0.9583 accuracy & 0.8684 F1 score.

VIEW CODE

Coinventor

• Collaborated on developing an app that converts from one currency to another with the current rate from Currency API.

• Managed the GitHub repository/submission process and designed a user-friendly UI following Material Design principles.

VIEW CODE

Lack of practicality in Vietnamese STEM Education

• Analyzed and ranked stakeholders and their factors that impede practicality in how STEM is currently educated at formal schooling in Vietnam in order of degrees of impact via an online survey.

📝 VIEW PAPER

Missing mass problem & the need for Dark matter (Lumeiere Research)

• Conducted a 15-page cosmological literature review that briefly introduces the ‘missing mass’ problems & explains why Cold Dark Matter is the most ideal theory mentored by Ph.D. student Matthew Scoggins (Grade A+).

• Selected as a Valedictorian speaker to present the final research paper live during the Lumiere Research Symposium for Spring 2023 Cohort.

📝 VIEW PAPER

Contact

(484) 588-8518

800 Lancaster Avenue, Villanova, PA 19085

hiepnguyenduc2005@gmail.com
hnguye30@villanova.edu