Experience
PricewaterhouseCoopers
- Leading cross-team collaboration to design & deploy generative AI models (including GPT-based LLMs) that automate complex regulatory tax data extraction & validation, reducing manual processing time & improving accuracy in compliance workflows.
- Engineered a full-stack AI platform using Streamlit & FastAPI, integrated with Azure Form Recognizer, cutting document processing time from 10 hours to 2 minutes & achieving 99% extraction accuracy across 10,000+ financial records.
- Improved model precision by 20% through demographic profiling & NLP techniques, boosting classification accuracy in automated tax form analysis.
- Built a Python-based chatbot and interactive dashboards (Tableau, Plotly) for a national nonprofit, increasing engagement by 70% among 1,200 students and improving service quality metrics by 35% through enhanced data-driven decision-making.
Jules Stein Eye Institute
- Published 7 papers in Ophthalmology & American Glaucoma Society journals, applying AI models to 3,000+ patient records to improve glaucoma progression forecasting.
- Built deep learning models in Python & R to predict visual field degradation, increasing early detection accuracy by 28%.
- Performed survival analysis & multivariate regression to estimate progression rates, uncovering 5 key prognostic factors used to tailor personalized care plans.
American Express
- Directed a team of 3 engineers to develop an ML pipeline processing 10.4M+ financial records, increasing classification accuracy by 30% & enhancing fraud detection precision.
- Analyzed campaign performance data using Pandas & Scikit-learn, optimizing resource targeting strategies & improving ROI by 80% across multiple campaigns.
- Deployed ensemble models in Jupyter Notebook to identify high-probability customer segments, increasing lead conversion by 3x.
- Automated reporting workflows, reducing model iteration cycle time by 50% & enabling faster stakeholder decision-making.