. Developed a CoPilot application in a Typescript-based monorepo that integrates Chat-GPT and Claude to assist clinicians in efficiently generating high-quality clinical notes, improving workflow and documentation accuracy.

. Refactored a Next.js full-stack architecture to utilize a Fastify-powered Node.js server that now handles up to 1M requests per month, enhancing performance and scalability within the Turborepo framework.

. Maintained and improved a BedBoard application, a patient lifecycle management tool that tracks admissions, discharges, room assignments, and authorizations across multiple facilities.

. Created real-time alerts for admissions teams, notifying them of available beds specific to each facility, streamlining patient flow and reducing response times.

. Developed automated data intelligence processes, including nightly emails with comprehensive census metrics (occupancy rates, average length of stays, discharge reasons etc.), reducing manual reporting by 75% and enabling leadership to make data-driven capacity decisions.

▪ Provide guidance and mentorship to high school students, fostering their understanding of fundamental machine learning principles, including data preprocessing, visualization, and model building.

▪ Worked one-on-one with each student to guide them through the process of conceiving, planning, and executing a capstone AI project.

▪ Designed predictive ML models to forecast conversion rates with a RMSE of 0.5%, a key metric for loan request approvals using python

▪ Implemented seasonality analysis for generating insights on loan request patterns using demographic, economic, geographic data to inform firms marketing strategy

▪ Leveraged a language model (LLM) to design and implement a prompt-driven tool, streamlining the extraction of valuable data from PDF documents used for pricing model

. Serving as a developer for a Computer Vision project aimed at localized identification of emission sources for air quality management

. Constructed a dataset identifying 500+ vehicles of 11 types by analyzing traffic surveillance feed from across the NY state

. Automated segmentation and tracking of vehicles by optimizing YOLO and SORT algorithms saving 6 min/detection

. Currently optimizing the code to handle occlusion issues and calibrate emission levels thereby increasing accuracy by ~9%

▪ Ensured seamless integration with third-party APIs for over 10 full-stack projects in collaboration with product owners, business and client teams, aiding critical lending solutions to banks.

▪ Worked on data extraction in loading JSON data from the server-side content into Salesforce database (RDBMS), demonstrating proficiency in crafting and optimizing SQL queries for streamlined data transfer.

▪ Tailored existing and built new dynamic lightning screens for Loan origination workflow engine using React and other front-end technologies with a high adoption rate of 90% among clients.

▪ Received “Prevailed Persistence Award” for contributions to the Closing Document integration, showcasing excellent stakeholder communication and customer support by assisting clients in setting up new release builds

▪ Developed a reusable and robust web-component, a JSON editor in Java, Javascript, React that significantly reduced request-response mapping time by 80%, enhancing operational efficiency.

▪ Built a visual analytics dashboard to monitor API calls using Lightning Frameworks, aiding key insights to performance.