Built an LLM-powered chatbot that enables internal teams to quickly find and validate information across thousands of health contracts, cutting contract review time by 80%.
Created a robust RAG system with custom document processing to improve retrieval and generation of relevant information from our knowledge base by 52%.
Developed an evaluation framework to test and improve AI systems, ensuring accurate, reliable responses for healthcare policy questions.
Deployed multiple Amazon SageMaker endpoints for Medicaid and other healthcare models, enabling batch and real-time inferences for data scientists and reducing manual reviews by 20%.
Streamlined ML infrastructure in AWS and Spark, reducing data engineering dependency by 40% and empowering 120+ non-technical users to self-serve insights through natural language to SQL querying.
Data Science Intern
ICURO
January 2022 - August 2022
Santa Clara, CA
Improved demand forecasting accuracy by 30% using advanced multi-step training techniques on large manufacturing datasets.
Built ML models in Azure ML Studio and PySpark to detect operational anomalies, reducing manual checks by 20% and preventing potential issues before they occurred.
Analyzed large-scale sensor and production data to identify supply chain inefficiencies, improving resource planning and operational efficiency.
Presented technical findings to business stakeholders, translating complex data into actionable recommendations.
Software Engineer Intern
Memorial Sloan Kettering Cancer Center
June 2019 - August 2019
Middletown, NJ
Built automated scripts that sped up patient data retrieval, helping doctors access critical information faster during treatment.
Collaborated with a senior developer to create backend endpoints that improved the performance of the OneMSK platform and streamlined access to essential resources.
Leading a team of four to build a habit-tracking app where users share daily photo proof of their progress, creating a social accountability system through personalized activity feeds.
Designed a scalable full-stack architecture that handles user interactions smoothly while preparing for future machine learning integration to validate habit completion photos.
Built the backend infrastructure for user authentication, media uploads, and habit tracking, ensuring a consistent experience across both Android and iOS platforms.
Rhythmia
Python, NumPy, Pandas, Scikit-Learn, Matplotlib, Spotify API
Developed a program that uses the athlete's current heart rate to find songs with a matching BPM, creating a personalized workout experience.
Integrated a K-Nearest Neighbors model to identify and queue similar tempo songs within 1 second of heart-rate input for seamless music transitions.
Automated dynamic playlist updates to help athletes maintain a consistent pace and reduce fatigue during workouts.
Applied research from Pearce et al. (2021) linking fast tempo music to the delayed onset of exercise-related mental fatigue.
Won the Health & Wellness track at HackRU 2022, competing among 120+ participants.
Created a recommendation engine that suggests songs based on user listening patterns and musical characteristics, helping users discover new music they'll enjoy.
Worked with Spotify's API to extract and process song metadata, analyzing features like tempo, energy, and mood to understand what makes each track unique.
Implemented a K-Nearest Neighbors model that learns from user preferences to provide personalized song recommendations, making music discovery more intuitive.
