Olivia Stefany

Machine Learning Researcher & Engineer

Representation Learning | Foundation Models | Speech AI | Scientific Machine Learning

Open to research collaborations

Get in Touch Request Resume

Olivia Stefany - Data Scientist

About Me

I'm passionate about bridging the gap between cutting-edge research and practical deployment. My research focuses on representation learning and foundation models for scientific and biomedical AI. Alongside academic research, I have built production machine learning systems in industry, shaping my interest in methods that are not only novel but also scalable, reproducible, and clinically meaningful.

With hands-on experience as an ML Engineer at Causalmind, I've built production systems using causal discovery and LLMs, deploying models on the cloud with Docker and managing complex data pipelines. This industry experience informs my research approach: I prioritize not just theoretical advances, but solutions that are interpretable, scalable, and clinically applicable. I believe the future of AI lies in creating foundation models that generalize across domains while remaining robust enough for healthcare and scientific applications.

Research & Publications

Research Interests

Representation Learning

Foundation Models

Speech AI

Scientific Machine Learning

Healthcare AI

Multimodal Learning

Manuscripts

Prosody-Conditioned Speech & Language Modeling

Submitted to EMNLP Status: Under Review

Focus: Speech representation learning for pathological speech processing

Speech AI Multimodal Learning

Speech-Based Machine Learning Models for Neurodegenerative Detection

Submitted to IEEE JBHI Status: Under Review

Focus: Hierarchical & multi-scale representation learning for clinical assessment

Representation Learning Healthcare AI

Ongoing Research

Audio Foundation Models

Status: Active Research

This project explores representation learning and multimodal alignment for audio foundation models with applications in scientific and healthcare domains.

Repository: Private (active research)

Efficient and Privacy-Preserving Hierarchical Federated Learning

Status: Active Research

Developing hierarchical federated learning systems that enable efficient collaboration across distributed institutions while maintaining strict privacy constraints through encryption and differential privacy techniques.

Focus Areas: Privacy-Preserving ML, Distributed Systems

Federated Learning Privacy-Preserving AI

AI Literacy and Competence Frameworks in Higher Education

Status: Active Research

Designing comprehensive frameworks for AI literacy and competence development in higher education, addressing how students and educators can effectively understand, evaluate, and responsibly deploy AI systems.

Focus Areas: Educational Technology, AI Ethics, Curriculum Design

Education Technology AI Ethics

Preprints

Coming soon

Publications

Coming soon

Experience

Research Assistant

Zhejiang University

July 2026 - Present

Built and designed foundational models and research prototypes for multimodal AI applications combining speech, text, and other modalities

Foundation Models Multimodal Learning Scientific AI Research Python

AI Researcher

Zhejiang University of Science and Technology

January 2026 - Present

Federated Learning & AI in Education

Contribute to improve federated learning systems that enable privacy-preserving AI training across distributed educational platforms without centralizing sensitive student data

Speech & Pathological Speech Processing

Design reproducible training, evaluation, and benchmarking workflows across multiple clinical speech datasets.
Implement data preprocessing, experiment management, and performance evaluation pipelines for research reproducibility.

Speech Processing Deep Learning Representation Learning Clinical AI TensorFlow/PyTorch Prosody Analysis Federated Learning Privacy-Preserving AI

Associate ML Engineer

Causalmind | Causal Discovery & AI Research

September 2025 - December 2025

Built an automated causal discovery system using Ollama LLMs and Neo4j, achieving 100% precision, 91.67% recall, and 95.65% F1 score for graph matching
Improved system reliability by designing loop prevention algorithms that reduced infinite loop occurrences by 90%
Engineered a hybrid causal discovery approach combining BIC scoring, PC algorithm, and LLM-based reasoning to enhance causal inference accuracy
Deployed containerized applications using Docker on cloud infrastructure to ensure scalability and high-performance reliability
Collaborated with stakeholders by producing detailed technical documentation and executive summaries that communicated business value to both technical and non-technical audiences
Learned production ML engineering practices including API design, semantic validation systems, and causal inference methodologies

Python LLM (Ollama) Neo4j Machine Learning Causal Inference Docker Azure REST API

Education

Bachelor of Science

Zhejiang University of Science and Technology

Major: Data Science and Big Data Technology

Expected Graduation: 2027

Projects

Featured Research

Advanced research projects on audio, speech, and multimodal AI

Audio Foundation Models visualization

Audio Foundation Models

Active Research

Building audio foundation model for scientific and clinical applications.

Active Development

Alzheimer's Speech AI visualization

Alzheimer's Speech AI

Under Review

Speech-based machine learning models for neurodegenerative disease detection using hierarchical representation learning.

Submitted to IEEE JBHI

Prosody-Conditioned Speech Modeling visualization

Prosody-Conditioned Speech Modeling

Under Review

Speech representation learning for pathological speech processing with prosody conditioning.

Submitted to EMNLP

Scientific Machine Learning

Deep learning for scientific discovery and healthcare applications

Brain CT scan showing tumor classification

Detect & Diagnose: Brain Tumor Classification with AI

Python TensorFlow CNN Deep Learning

A deep learning project using CNNs and transfer learning to classify brain CT scans, identifying tumor types or normal brains, helping accelerate and improve diagnostic accuracy.

QSAR / ADMET Prediction visualization

QSAR / ADMET Prediction for Drug Discovery

Python Machine Learning Drug Discovery

Quantitative Structure-Activity Relationship modeling for drug discovery with ADMET property prediction and therapeutic target evaluation.

View Documentation

Applied Machine Learning

Production ML systems for real-world applications

Smart Finance System

Smart Finance Budgeting System

Python Time Series Forecasting Streamlit Prophet Plotly

Built a multi-user personal finance dashboard with bcrypt auth, SQLite storage, 30-day Prophet spending forecasts, budgeting insights, and multi-currency conversion (real-time + offline). Built with Streamlit + Plotly.

GitHub Live Demo

Credit card approval prediction visualization

Predicting Credit Card Approval

Python Scikit-learn Classification

Machine learning models to predict credit card approval based on applicant data such as income, credit score, and employment history, helping automate decisions and improve accuracy.

Housing market price prediction chart

Unlocking Housing Market Insights with Linear Regression

Python Regression Data Analysis

Implemented Linear Regression model to forecast house prices, emphasizing data preparation, and feature evaluation and model performance metrics.

Skills & Expertise

Research Areas

Representation Learning Foundation Models Speech Processing Healthcare AI Scientific Machine Learning

ML & Deep Learning

PyTorch TensorFlow Hugging Face Scikit-learn LightGBM XGBoost

Programming & Tools

Python Docker Git SQL Neo4j REST APIs

Specialized Skills

Speech Analysis Clinical AI Federated Learning Causal Inference Time Series Forecasting Computer Vision

Certifications

Data Scientist

Kelas.com

September 2025

Certificate ID: CERT-13B86355

Verify Certificate

Let's Connect

Interested in discussing opportunities, collaborations, or just want to chat about data science and AI? I'd love to hear from you!

I typically respond within 24 hours

Schedule a Meeting

Contact Me

me@oliviastefany.dev | (+86)18668233076

Currently based in China | Open to remote opportunities

Email Me