Data Science @ Metropolitan State University | GPA: 4.0 | Two-Time 1st Place State Winner
I find the blind spots in systems and build working tools to fix them — healthcare access gaps, unfair team dynamics, unverified contributions. Not just analysis. Deployed products people can use.
I designed a three-app product loop that covers the full lifecycle of a team:
FORM a team → RUN it healthily → PROVE who contributed
| App | What it does | Live demo |
|---|---|---|
| TeamUp | Skill-based team matching + kickoff | Open app |
| Team Doctor | Charter, RACI health checks, EOS meeting rhythm, anonymous pulse + coach | Open app |
| Clearwork | Log deliverables, peer confirmation, balance analysis, verified resume cards | Open app |
Why this exists: I watched teams fall apart — not from conflict, but from invisible problems: unclear ownership, uneven workload, no structure. These apps make that structure explicit and keep it honest.
How it's built: Pure Python + Streamlit. All health checks are deterministic — no LLM, no API key, no cost, nothing that can hallucinate. Append-only decision records. Structurally anonymous pulse surveys. Schema-versioned entities for future migration.
Built without AI — pure HTML/CSS/JS solving my own daily computer-time pain points. Browser-based, no sign-up, no backend, installable as PWAs. Deployed on Cloudflare Pages.
| App | What it does | Live demo |
|---|---|---|
| FocusStone | Pomodoro timer with 12 guided stretches, voice reminders, break planner, optional ESP32 physical button | Open app |
| FlowTime | Adaptive focus timer — schedules work/rest around cognitive load, not fixed 25-min blocks. Classroom mode for instructors | Open app |
| FlowMove | Turns every break into a guided stretch/eye-care routine matched to your sore areas + break length. Curated video demos + offline timer | Open app |
Self-contained offline study sites I built for my statistics coursework — concept explorers, decision trees, comparison tables, and interactive graphs. No frameworks, no backend.
| Course | What it covers | Live site |
|---|---|---|
| STAT 353 | SPSS methods — hypothesis testing, ANOVA, regression, chi-square. Interactive concept explorer + method decision tree + SPSS step-by-step guides | Open site |
| STAT 301 | R-based statistics — linear algebra, calculus review, LDA, gradient boosting, hypothesis testing. Same interactive framework, R instead of SPSS | Open site |
- 1st Place — HeatMap Hackathon (03/2026): Ranked #1 in the state for identifying geographic disparities in U.S. burn care access. Built the financial and geospatial engine — modeled state-level avoidable hospitalization costs, scored 498+ trauma centers as telemedicine hub candidates using Census and NIRD 2023 data.
- 1st Place — Minnesota AGENT.AI Competition (10/2025): Ranked #1 in the state for predictive accuracy in retail banking fraud detection. Shifted from baseline accuracy to behavioral/temporal features with imbalance-aware learning (Macro-F1: 0.5211 fraud, 0.9553 default).
- Dean's List x 6 semesters & 5 Scholarship Recipient
| Project | What I did |
|---|---|
| Burn-Care-Equity-ROI | Geospatial ROI engine behind the 1st place hackathon win — Folium mapping, Tele-Burn Hub scoring, cost modeling |
| AgentDS-Retail-Banking | Domain-driven fraud detection pipeline behind the 1st place AGENT.AI win |
| ZeroCode-Career-Coach | Redesigned a 7-agent pipeline into 3 LLM agents + deterministic orchestrator. ATS scoring, resume tailoring, voice mock interviews. Runs fully local on Ollama |
| JiuZuoShu (久坐舒) | TCM self-care app for desk workers — 9-constitution quiz, personalized daily routines, stretch reminders. Zero-cost, no API key |
| Developer-Skills-Analysis | IBM Capstone — 70k+ record pipeline analyzing tech trends with Python, SQL, and Power BI |
Languages: Python, R, SQL, Java, JavaScript, HTML/CSS Data: Pandas, NumPy, Scikit-learn, TensorFlow, JMP, SPSS Visualization: Power BI, Tableau, Streamlit, Folium, Matplotlib, Seaborn Cloud & Infra: Snowflake, Databricks, AWS (EC2, S3), Docker, Git AI/ML: LLM orchestration (Ollama), multi-agent systems, NLP, computer vision (OpenCV, YOLO, LLaVA), prompt engineering
Let's connect: LinkedIn | Portfolio | feifeiusa30@gmail.com