• Developed end-to-end development of real-time Computer Vision system using PyTorch and YOLOv8 for object detection,achieving 91% accuracy and 30% performance improvement through optimized image pre-processing pipeline• Implemented LOFTR matcher with PyTorch for damage tracking, utilizing Homography matrix for viewpoint invariance and achieving 91% accuracy through RANSAC-based outlier filtering• Fine-tuned YOLOv8 model on custom dataset of 5,000+ images using PyTorch, improving object detection performance by 40% through transfer learning and data augmentation tracking of pre-existing damages.• Established automated model monitoring using Weights & Biases and MLflow, reducing model error by 15% through systematic experimentation and achieving 91% precision in production

Assigned a project at Althea Health Inc. to address the need for enriching doctor-patient communication duringmedical appointments.• Crafted a specialized Conversational AI model to forecast turn-taking in medical consultations, striving for moreorganic interactions.• Utilized advanced Natural Language Processing (NLP) techniques and machine learning algorithms to architect,refine, and assess AI frameworks like GPT-2 and DistilBERT for medical dialogue prediction.• Implemented AI model that significantly improved communication dynamics in medical consultations, achieving anaccuracy of 95.81% with GPT-2 and 95.69% with DistilBERT, showcasing its efficacy in augmenting doctor-patientcommunication within healthcare settings

• Built and deployed personalized recommendation models using XGBoost and collaborative filtering across 15 partner colleges, improving prediction accuracy by 25% and serving 500+ students per semester.• Designed end-to-end recommendation system processing data from 8,000+ users, combining Python/SQL featureengineering with PyTorch models, achieving 60% performance improvement and 18% higher user engagement across 20+ institutions.• Led development of real-time monitoring systems for recommendation performance, analyzing behavior patterns and interaction metrics for 12,000+ users through SQL-based analytics pipelines• Scaled recommendation infrastructure using PySpark and AWS, optimizing pipeline by 40% for millions of daily interactions while implementing MLflow for production model management•