AI Workflow & Visual Data Projects
Student projects reframed for AI workflows and infographic-style storytelling, with clear structure for quick scanning.
Explainable Lymphography Diagnosis Workflow (XAI)
Summary: Transformed a medical classification model into a visual explanation workflow that helps users understand predictions.
Problem: Medical predictions can be accurate, but users may not trust them when model reasoning is not visible.
Approach: Designed a full evaluation pipeline using KNN, Decision Tree, and Random Forest, then analyzed feature influence with SHAP to understand model behavior beyond accuracy.
Visual Output: Visualized SHAP feature impact, class-level comparison, and prediction reasoning in structured chart sections so non-experts can read "what influenced this output" quickly.
Outcome / Impact: Improved interpretability and evaluation confidence by converting black-box predictions into understandable visual insights for student research use.
Data Story Flow: Raw medical features -> model predictions -> SHAP explanations -> clear insight cards.
Role Alignment Upgrade: Add an interactive dashboard with patient-level SHAP views, model comparison charts, and a one-page infographic summary.
Tools: Python, Scikit-learn, SHAP, Matplotlib
Daily Sugar Guidance: Data-Driven Decision Support App
Summary: Converted daily sugar readings into clear visual guidance blocks for faster health-related decisions.
Problem: Users often see sugar numbers but struggle to decide the right next action in real time.
Approach: Analyzed reading ranges, designed rule-based thresholds, and built a Streamlit interface with input validation and edge-case handling.
Visual Output: Structured recommendations into readable output blocks (diet, walk, alert) with clear priority and clean language for easy understanding.
Outcome / Impact: Increased readability and practical use by transforming raw readings into clear, immediate actions.
Data Story Flow: Raw sugar values -> risk banding logic -> action-focused visual guidance.
Role Alignment Upgrade: Add trend charts, color-coded threshold bars, and a simple infographic panel showing "today vs safe range".
Tools: Python, Streamlit, GitHub, Streamlit Cloud
Portfolio Information Design for Data Storytelling
Summary: Redesigned project communication into a structured, visual-first format for clearer story flow.
Problem: Generic portfolio pages can hide value when project information is dense or unstructured.
Approach: Analyzed content hierarchy, redesigned section flow, and rewrote project narratives using consistent structure and strong action verbs.
Visual Output: Presented information in card-based sections with clear headings, compact paragraphs, and labeled insight blocks for better scanability.
Outcome / Impact: Improved clarity and recruiter readability, making project strengths easier to understand in less time.
Data Story Flow: Raw project details -> key insight selection -> structured visual narrative.
Role Alignment Upgrade: Add before/after wireframe visuals and a mini infographic showing improvements in readability and information hierarchy.
Tools: HTML, CSS, Content Structuring, Visual Hierarchy