Responsibilities include:

• Developed a CNN-based segmentation model to measure vertebral motion for assessing fusion status in ACDF surgery using labeled cervical radiographs.
• Achieved 99.2% accuracy in spinous process detection by utilizing data augmentation techniques including noise stimulation, blurring, cropping, and rotating.
• Calculated ISPM by converting segmented radiographs into BLOB images, extracting rightmost spinous process coordinates, and measuring pixel distance between them.

Responsibilities include:

• Collaborated with Kheyti, a nonprofit in India, to develop a crop planning decision support tool for small farmers using Multi-Agent Reinforcement Learning.
• Implemented the Multi-Agent Rollout policy that simplifies the global optimization problem into sequential local decisions, leading to increased fairness and overall outcomes for farmers.
• Designed a MySQL database schema to manage simulation data and used Matplotlib and Seaborn to visualize the effectiveness of policies in real-world agricultural settings.

Responsibilities include:

• Developed surrogate models for energy-efficient building design, providing urban planners with real-time decision-making tools to support sustainable city planning in Atlanta.
• Led an interdisciplinary subteam to design a multivariate regression model that reduced simulation times from hours to seconds, utilizing Gradient Boost and Bayesian Optimization.
• Created a game-based tool designed to teach surrogate modeling, allowing architects to optimize building design parameters through interactive challenges aimed at minimizing energy consumption.

Responsibilities include:

• Collaborated with Emory’s Morningside Center to develop an intuitive clinician interface that accelerates quantitative meta-analyses for clinical trial literature review, focused on drug repurposing.
• Proposed and implemented a full-stack web application with Next.js, incorporating interactive Plotly visualizations and user-access management through AWS sign-in authentication.
• Utilized Lambda Functions and OpenAI Gateway to develop advanced filtering capabilities, including a PubMed Query Generator and customizable results table, processing over 1,000 studies in under a minute.

Responsibilities include:

• Collaborated with Merck’s R&D department to fine-tune a LLM for automating marketing content audits, ensuring the model accurately flagged content violating nuanced regulatory guidelines
• Employed prompt engineering and developed a curated dataset of industry-specific compliance use cases, resulting in a 20% improvement in model accuracy.
• Built an AI chatbot using Langchain, ChatGPT API, and Hugging Face, streamlining compliance workflows and enabling peer collaboration on prompt engineering and use-case curation.