Shusei Yokoi

Bias by Prompt LLM Fairness | Python, ChatGPT, Claude, Gemini, Qwen

• Conducting research on trustworthy AI, focusing on fairness, bias, and reliability in LLM-based decision-making
• Designed controlled experiments to evaluate the effects of prompt framing on responses across LLMs, including ChatGPT, Claude, Gemini, and Qwen
• Identified early evidence that emotional prompt framing can alter LLM behavior, causing conclusion shifts in some models and reduced confidence in others

Founded and Developed an AI-powered Solutions | R, Python, Swift, Azure OpenAI

• Founded Datify to build AI-powered products that help people improve their lives through personalization
• Developed HealthSync, an iOS AI health advisor app integrating Apple HealthKit and Azure OpenAI GPT-4o to deliver personalized health insights from biometric and activity data
• Built data pipelines to extract biometric data from Apple HealthKit and convert them into structured summaries for AI analysis to generate dynamic, contextual health advice, helping users track trends and stay on target with personal goals

Gym Chain Health Data Analysis | R, SQL

• Analyzed 200,000 member records for a gymnasium company with 150+ branches across Japan.
• Identified a 3-month weight regain trend in younger members and developed tailored retention strategies.
• Implemented a notification service to re-engage inactive members, increasing re-engagement rate by 30%.

Software Development Team Productivity Analysis | SQL, Python, R

• Performed Difference-in-Differences analysis on ticketing system data to diagnose productivity bottlenecks.
• Found resource allocation inefficiencies and recommended more frequent ticket creation and resource optimization.

Led Application Development | AWS, Azure, JavaScript, GitLab, SQL, VoltMX

• Directed end-to-end development of a multi-platform office management system for teams in Vietnam, China, and Japan.
• Led UI/UX design, back-end architecture, testing, and cross-platform deployment.

Trade Area / Population Flow Analysis | Tableau, SQL, Python, R

• Led trade area analysis for Izumi Co. with 190+ malls under SoftBank's Smart City project.
• Built Tableau dashboards using GPS, demographic, and search data to uncover customer trends.
• Recommended targeted ads, in-store improvements, and loyalty strategies based on retention and regional growth insights.

AI Engineer | Python, SQL

• Developed a population inflow prediction model to optimize billboard advertising placement.
• Improved model performance through data engineering and feature design, achieving an AUC of 0.70.

EC Site Analysis | SQL, Python, R, Google Analytics

• Analyzed EC site traffic using Google Analytics and modeled sales patterns across product categories.
• Predicted product sales using a multilevel model with category-specific price elasticity.

AI & Machine Learning Research

LLM evaluation and sycophancy testing across GPT, Claude, and Gemini; algorithmic bias auditing and mitigation using SHAP and XGBoost on real-world lending data; supervised learning, deep learning, and statistical hypothesis testing.

Winning award at USC AI4SE Hackathon 2026

Won the Participants' Choice Award for CastIt, an AI podcast platform with a React/TypeScript/Supabase frontend that turns news articles into personalized audio episodes via an LLM-driven pipeline. Built the frontend end-to-end while a teammate built the backend pipeline.

Data Science & Statistics

Statistical learning, regression analysis, multilevel and mixed modeling, categorical data analysis, statistical computing in R, time series, forecasting, and model evaluation.

Programming & Systems

Python application development, database systems, ERD/UML, advanced SQL, systems analysis, SDLC, UI/UX requirements, project management, and blockchain development.