Mia Zhou

profile

Mia Zhou

Mathematics (B.S.) & Biostatistics (B.S.P.H) · Minor in Neuroscience
University of North Carolina at Chapel Hill · Expected May 2027

Email Phone LinkedIn GitHub Scholar

// philosophy

Research has taught me that the most valuable ideas emerge when curiosity meets execution. As an aspiring AI and Data Scientist, I enjoy exploring new questions, learning across disciplines, and transforming complex data into actionable insights. My current research focuses on AI applications and security, biostatistics, network science, and survival analysis.

In the long term, I hope to become both a researcher and an entrepreneur — creating AI technologies that move beyond publications to deliver lasting real-world impact. I am driven by continuous growth, interdisciplinary collaboration, and the belief that thoughtful innovation can improve how people live, learn, and make decisions.

internship / work experience

Jun – Aug 2027

Incoming Business Analyst Intern

Capital One · McLean, VA

Jun — Aug 2027

Incoming Business Analyst Intern

Slated to engineer data-driven product strategies leveraging cloud-native financial data infrastructure.
Will utilize SQL and Python (Pandas) to query AWS Snowflake data warehouses and model user behavioral KPIs.
Prepared to translate machine learning insights and A/B test metrics into technical Agile/Jira backlogs.

Jul – Aug 2026

Software Engineer Intern

Google Raleigh-Durham · Break Through Tech Sprinternship

Jul — Aug 2026

Software Engineer Intern

Led a 4-person Agile team to develop an agentic AI application using Python, React/Vite, Google BigQuery (SQL), Gemini API, and GCP, analyzing 41 GTFS transit feeds to study educational accessibility for 2,400+ public schools across 100 counties.
Designed a RAG LLM workflow using the Google ADK framework and Gemini to generate policy recommendations, paired with logistic regression risk-factor analysis.
Built EduBridge, an application exploring factors contributing to educational disparities across North Carolina.

publications

Shi, C.*, Han, S., Kusumanchi, S., Zhou, M., & Li, D. NIH-MPINet: A Large-Scale Feature-Rich Network Dataset for Mapping the Frontiers of Team Science. SLADS. Accepted

Gao, Y.*, Chen, L.*, Wu, W.*, Zhou, M., Wang, J., Ji, K., Guo, M., & Chen, H. (2026). Bit-flip attacks on vision-language-action models: Action-decoding architecture shapes the vulnerability. AAAI 2027 Conference Submission. Submitted

Gao, Y.*, Chen, L.*, Wu, W.*, Zhou, M., Mao, X., Guo, M., & Chen, H. “Cooldown Landmines: Cross-Tenant Denial of Service via Shared Circuit-Breaker State in LLM Gateways.” ICLR 2027 Conference Submission. Submitted

research experience

Mar 2026 – Present

Multi-State CVD Progression Modeling

Honors Thesis · UNC Chapel Hill — Dept. of Biostatistics

Honorable Mention @ BUSI Symposium 2026

Mar 2026 — Present

Multi-State CVD Progression Modeling

Built a 6-state disease progression model across ARIC, CHS and FHS, generating each 10-year cardiovascular risk predictions.
Automated quality-control workflows, resolving sparse-data bias and hidden preprocessing errors across 20 model evaluations.
Validated predictive performance using discrimination C-index and Calibration, achieving 78.5% HF discrimination and 82.7–87.2% for multi-stage disease prediction.
Presented the poster at Biostatistic Undergraduate Summer Internship Symposium and received Honorable Mention.
Writing manuscript and extending the thesis to develop and optimize survival prediction methods by integrating existing statistical approaches.

Sep 2026 – Present

Quantitative Research Intern

UNC Undergraduate Research Training Group

Sep 2026 — Present

Quantitative Research Intern

Developing robust optimization models for bond portfolio construction using alpha signals and market data.
Building automated data, modeling, and backtesting workflows to evaluate portfolio strategies, risk, and performance.

Oct 2024 – Present

Project 1 — NIH Network Analysis

Project Assistant · UNC Biostatistics — Li Lab

Oct 2024 — Present

Project 1 — NIH Network Analysis

Engineered an NLP pipeline using BERTopic and BioSentVec to analyse 30k+ NIH grants and identify funding trends.

Code

Oct 2024 – Present

Project 2 — Spatial Transcriptomics

Project Assistant · UNC Biostatistics — Li Lab

Oct 2024 — Present

Project 2 — Spatial Transcriptomics

Benchmarking 3 graph-learning frameworks (Seurat, Squidpy, Niches) for spatial transcriptomics, on Stimage-1K4M dataset to evaluate computational efficiency and downstream biological performance.

Jun 2024 – Present

eNet Enhancer Network

Research Intern · Xiamen University — Huang Lab

Jun 2024 — Present

eNet Enhancer Network

Integrated single-cell transcriptomic and epigenomic datasets (18K+ cells) to construct graph-based enhancer networks and analyze gene regulatory relationships in leukemia.
Applied network analysis, correlation modeling, and co-accessibility metrics to identify regulatory modules for the Leukemia driver gene.
Analyzed enhancer topology and interaction patterns, discovering that modular enhancer networks (~8.4%) exhibited significantly stronger regulatory enrichment than non-modular networks.

Report

May – Aug 2025

Agent-Based Economic Simulation

Financial AI Intern · Tsinghua University

May — Aug 2025

Agent-Based Economic Simulation

Built infra for an LLM-powered Agent Society, configuring remote Linux environments, SSH port forwarding, dependency management, and experiment workflows for large-scale simulations.
Customized prompts and structured outputs for 100 autonomous AI agents operating in a virtual Beijing environment, enabling coordinated economic decision-making within a behavioral RCK model.
Executed and analyzed reproducible multi-agent simulations, evaluating emergent business cycles, wealth inequality, and belief polarization using experiment tracking and sensitivity analysis.

PDF Code

competitions & hackathons

Mar 2026

Duke ASA DataFest 2026Best Statistical Analysis

Duke University · American Statistical Association · 5-person team

Mar 2026

Duke ASA DataFest 2026

Led EDA cleaning pipeline for a 4-person team using Python, R, Scikit-learn on 7M+ healthcare data over 8 social risk domains, and won the Best Statistical Analysis Award.
Identified high-risk diabetic populations using logistic regression, projecting an 18% reduction in avoidable ED visits.
Designed nearest-neighbor mobile clinic routing covering 32K+ high-risk patients within a 15-mile radius and built an interactive geospatial dashboard for healthcare planning.

Slides Code Map

Mar 2026

WageGuard — SolHacks 2026 (MLH)1st Place Culture Track

Product Manager + Full Stack Developer · 2-person team

Mar 2026

WageGuard

Led product design and front-end development for a bilingual, voice-first wage protection platform helping Latino workers identify wage theft, understand labor rights, and generate actionable complaints.
Built a 3-agent AI workflow using Gemini and ElevenLabs to convert natural-language voice input into structured wage claims, ask adaptive follow-up questions, and generate plain-language English or Spanish guidance.
Engineered an Express/TypeScript backend with Supabase authentication and a deterministic 50-state wage calculation engine, separating legal calculations from LLM reasoning to improve accuracy and reliability.
Won 1st Place at SolHacks after delivering the end-to-end product in 24 hours, including retaliation detection, evidence organization, downloadable case records, and complaint generation.

Devpost

Apr 2026

Duke Data Mini #3Best Statistical Analysis

Duke University · Railroad fatality modeling datathon

Apr 2026

Duke Data Mini #3

Built a two-model risk-analysis pipeline combining logistic regression and XGBoost on U.S. railroad incident records to identify fatality risk factors.
Found trespassing (33.85x odds) and positive drug tests (9.53x odds) were the dominant mortality drivers, explaining the bulk of the fatality gap.
Clustered fatal incidents into recurring behavioral archetypes (track-walkers, grade-crossing drivers, maintenance-zone workers) to point toward targeted interventions.

Slides

projects

NC State Government Solveathon 2026

Sponsored by NC Office of State Budget and Management (OSBM) · Client: NC DNCR

NC State Government Solveathon 2026

Co-developed an Azure-based AI platform for NC government, reducing 1,364 federal listings to 15 personalized recommendations (98.9% less to review).
Built a Python machine learning pipeline (scikit-learn, TF-IDF) aligned with 25 strategic objectives; benchmarked 3 LLMs on 50 grants and helped test 2,556 scoring configurations, reaching 80% precision.
Modeled 257 staff hours saved per year (87%) and a $1M increase in expected grant funding through a one-year simulation.

Demo Code

HerTime

Data Scientist & Analyst · Carolina Data Challenge 2026

HerTime

Built a data pipeline in Python (pandas, NumPy) merging 7 World Bank datasets (129K+ records, 215 countries), and rebuilt missing rural-women data at r = 0.998 against 1,263 direct surveys.
Modeled gender opportunity gaps with a Bayesian model (Gibbs sampling, Kalman filter), finding rural women's jobs trail urban women's by a median of 21 years.
Built a LightGBM + SHAP early-warning predictive model predicting widening gaps at 0.87 AUC (vs 0.54 baseline), and delivered 2,050 findings to a LangGraph multi-agent AI assistant on the website.

Live Demo Code

Jan – May 2026

Clinical Doppelgänger — EHR Clinical Decision Support

Full Stack Developer · CHIP 741

Jan — May 2026

Clinical Doppelgänger

Built a SMART on FHIR web app that securely retrieves and continuously updates 5 types of patient health records, turning them into an interactive clinical timeline.
Developed a patient-matching pipeline using Propensity Score Matching, Kaplan–Meier analysis, and bootstrap CIs to identify similar patients and compare their outcomes over time.
Trained ML/predictive models on SMART sandbox patient data to generate personalized patient trajectories, combining historical "patients like me" patterns with continuously updated backend data for clinical decision support.

Live Demo Code Slides

Jun – Aug 2025

Neuromatch Neuron Simulation

NeuroMatch Computational Neuroscience Program

Jun — Aug 2025

Neuromatch Neuron Simulation

Analyzed large-scale IBL neural recording datasets to investigate thalamocortical communication flow and decision-making dynamics during visually guided behavioral tasks using neural decoding and latency analysis.
Built machine learning pipelines using pre-stimulus neural activity to decode internal bias and choice behavior, showing higher decoding accuracy under zero-contrast conditions.
Applied cross-correlation, PSTH analysis, PCA, and ensemble classifiers (Random Forest, XGBoost, SVM) to identify temporal hierarchy and region-specific neural representations.

Slides

leadership / volunteer / honors

UNC-Chapel Hill Women in Data Science

Founder & President · Women in Data Science Ambassador, Worldwide WiDS Community

2026 — Present

Founded and lead the UNC-Chapel Hill WiDS chapter while serving as a Women in Data Science Ambassador in the worldwide WiDS community.
Building an AI + data space for the next generation of students through workshops, speaker series, hackathon programming, and interdisciplinary data-science events.
Planning UNC-hosted workshops, hackathon programming, and speaker series connecting statistics, AI, health, and social impact.

MATH 523 Complex Analysis

Undergraduate Learning Assistant · UNC Chapel Hill — Department of Mathematics

Present

UNC Biostatistics Program

Ambassador · Bios BSPH program