I believe in the power of community and collaboration through data to instigate positive societal change. My background in psychology and sociocultural development, combined with industry experience creating multidisciplinary tech solutions, gives me a unique perspective at the intersection of data science and human factors.
I've implemented machine learning solutions across multiple domains (data privacy, education, healthcare, and design) with a focus on ethical AI use and digital transparency that promotes mental health and safety.
👩🏽💻 MS in Data Science from University of San Francisco
👩🏽🎓 BA in Data Science and Psychology from University of California, Berkeley
📝 Creative writer at Substack
🎨 Artist/designer at Instagram
- Data Scientist at Buck Institute for Research on Aging
- Data Scientist at Tandem Diabetes Care
- Research Fellow at UC Berkeley School of Social Work (HERE2PSW)
Languages & Tools:
ML & Data Science:
Data Engineering & MLOps:
Cloud & Platforms:
Design & Productivity:
Privacy Policy Analysis Tool
- Built NLP-powered Chrome extension identifying harmful privacy clauses
- 95% reduction in manual review time
- Tech: Python, JavaScript, HTML/CSS, LLM, Chrome APIs
- 🏅 Award: 1st Place (audience vote) and 2nd place (judges vote) at USF Entrepreneurship Summit
Behavioral Pattern Analysis for Student Well-being
- Analyzed 11,620+ schools across 2,245 safety attributes
- 45% improvement in analysis efficiency through automation
- Tech: Python, Pandas, SPSS, Scikit-learn, Statistical Analysis
ML-based Music Composition Tool
- Decision tree model trained on 1,036 MIDI files
- Real-time chord generation with Streamlit interface
- Tech: Python, Hugging Face, librosa, Streamlit
Generative Music Exploration App
- Built custom transformer architecture for symbolic music generation with 31,034 MIDI files from Lakh dataset
- Achieved 2.43 validation perplexity through 8-layer decoder model with 12-head attention and 768-dimensional embeddings
- Implemented genre/artist/era conditioning system supporting 15 genres, 2,956 artists, and 60 years of musical styles
- Tech: PyTorch, Transformer Architecture, MIDI Processing, Music Information Retrieval
Post-Referral Patient Support Tool
- 100% accuracy using MedDialog dataset
- Responsive web interface for patient education
- Tech: Anthropic API, JavaScript, HTML/CSS
Deep Learning for Biological Research
- 20% accuracy improvement through advanced preprocessing
- PyTorch & MONAI-based ResNet50 implementation
- Tech: PyTorch, MONAI, Computer Vision, Feature Engineering
Healthcare Data Analysis Platform
- Analyzed 6,573+ participant clinical trial data
- Predictive modeling for treatment optimization
- Tech: SQL, Databricks, Statistical Modeling, Visualization
- 11,620+ schools analyzed for youth safety patterns
- 95% reduction in manual privacy policy review
- 45% improvement in data analysis efficiency
- 20% ML model accuracy improvement through optimization
- 6,573+ clinical trial participants supported through analytics