Research-Oriented Computer Science · Human-Centered Data Systems

Exploring how data, machine learning, and design frameworks support human decision-making under uncertainty.

I am interested in how people make decisions in complex, information-dense environments, and how data-driven systems can be designed to reduce cognitive overload rather than amplify it.

My work explores:

• how raw data can be structured into decision frameworks
• how machine learning models can remain interpretable and trustworthy
• how visualization and interaction design influence sensemaking

Rather than optimizing models in isolation, I focus on human-centered data systems that support clarity, comparison, and reflection.

• Decision-making under uncertainty
• Information overload and cognitive framing
• Explainable and interpretable machine learning
• Data-driven sensemaking tools
• Human–AI interaction
• Generative AI for analysis and storytelling

M.S. in Computer Science
Northeastern University, Miami

Background in data science, machine learning, and system design, with a strong interest in research-oriented and design-integrated applications of AI.

Repository: https://github.com/NoMooncake/mira-astrology-companion

A large-scale, cloud-native AI system exploring how LLM-based agents can support emotional reflection and decision-making, rather than task completion.

Research focus

• Human-centered conversational AI
• Explainable emotional risk flagging
• Reflection over optimization
• System-level tradeoffs (cost, latency, interpretability)

Highlights

• Serverless architecture (AWS Lambda, API Gateway, DynamoDB, Bedrock)
• Rule-based and LLM hybrid reasoning
• Cost-sensitive design decisions
• Synthetic-data-first research ethics

Mira functions as a living research system, integrating architecture, ML reasoning, and human-centered design.

Repository: https://github.com/NoMooncake/behavior-state-modeling

A research-oriented framework for modeling and inferring latent user behavioral states from interaction patterns, designed to support downstream decision-support and adaptive systems.

The project translates low-level behavioral signals into interpretable, high-level user states, emphasizing structure and explainability over black-box prediction.

Research focus

• Behavior-to-state abstraction
• Interpretable user state modeling
• Temporal patterns and session-level signals
• Foundations for adaptive decision-support systems

This project serves as a foundational modeling layer for understanding user behavior prior to intervention, personalization, or automation.

Demo: https://nomooncake.github.io/YueWu3160

A data-driven study examining how attention and trends form across TikTok and YouTube, and how social platforms structure the information environments in which users make decisions.

Repository: https://github.com/NoMooncake/wine-bad-data-robustness

An empirical study on how machine learning models degrade under label noise, missing values, and corrupted data, focusing on robustness rather than peak accuracy.

Research focus

• Failure modes of ML models
• Robust evaluation under non-ideal data
• Comparative behavior across model families

Repository: https://github.com/NoMooncake/mini-vibes

A lightweight, rule-based prototype for emotion-aware risk flagging in short journal entries, designed for interpretability over prediction.

Research focus

• Explainability-first AI
• Conservative, non-diagnostic design
• Human-in-the-loop decision support

Repository: https://github.com/NoMooncake/mini-autograd

A minimal autograd engine built from scratch to study backpropagation, computational graphs, and learning dynamics.

Research focus

• Understanding learning mechanisms
• Transparency in optimization
• Educational and interpretive ML tooling

Repository: https://github.com/NoMooncake/meal-planner

A Java-based system modeling everyday decision-making under constraints, combining pantry state, preferences, and recipe logic.

Research focus

• Decision modeling in daily life
• Constraint satisfaction versus recommendation
• Modular system design

Data and Machine Learning

• Statistical modeling
• Interpretable machine learning
• Feature analysis and evaluation
• ETL pipelines and exploratory analysis

Systems and Design

• Modular system architecture
• MVC and design patterns
• Decision framework design
• Visualization and narrative structure

Technical Stack
Python · Java · SQL · Pandas · NumPy · scikit-learn · PyTorch
Tableau · Docker · AWS · Git · Jupyter · Maven

• Designing decision-support systems rather than prediction-only models
• Exploring generative AI as a co-analyst and explanatory agent
• Studying how different data representations influence user judgment
• Bridging machine learning and interaction design

• How much model accuracy should be sacrificed for interpretability?
• When does more information reduce decision quality?
• Can AI systems explain why not instead of only why?
• How can data systems support reflection rather than speed?

Portfolio: https://NoMooncake.github.io
LinkedIn: https://www.linkedin.com/in/yue-wu-030816d
Email: wu.y26@northeastern.edu

This GitHub serves as a living research notebook rather than a showcase of finished products.