• Implemented a Sales AI system using CrewAI, LangChain, and the Gemini model to build agents that send personalized emails to clients about products. Enhanced client engagement and lead conversion by developing reply agents with in-depth product knowledge to respond to client queries and assist in scheduling meetings.• Developed an OCR solution utilizing Google's OCR technology to convert TIFF files into PDF format, streamlining the conversion of bills into readable and manageable files. Improved organizational efficiency in handling financial documents by automating and simplifying the billing process.• Created agents using CrewAI and the Gemini model to generate LinkedIn posts based on specified topics, mimicking the writing style of the designated account. Streamlined content creation and maintained a consistent online presence, enhancing brand consistency and audience engagement.• Developed a comprehensive WhatsApp chatbot using Gemini that goes beyond templated responses to provide detailed and accurate answers to customer queries. Enhanced user interaction and support by delivering a more personalized and informative customer service experience.

• Implemented a 3D lung segmentation model using nnU-Net to classify lung regions as left lung, right lung, or trachea. Achieved a 97% accuracy, demonstrating advanced skills in medical image segmentation and deep learning.• Developed a 3D liver delineation model using nnU-Net to identify different liver segments from NIfTI files. Involved detailed analysis to understand various liver parts for tumor detection, resulting in a 98% accuracy.• Designed and trained a custom CNN for classifying kidney stones as either inside or outside the urethra. Extracted features from the CNN and trained a decision tree classifier, achieving a 90% accuracy.

SBERT Model Implementation and Cloud Deployment • Developed and pre-processed a dataset of 200 texts with labels for multi-label text classification, ensuring optimal formatting for model training.• Trained SBERT to generate fixed-length vector encodings, capturing semantic meaning for diverse NLP tasks. Utilized U-Net architecture for segmentation tasks, showcasing expertise in medical imaging analysis.• Deployed SBERT model as a REST API using Flask and Docker, enabling scalable and efficient sentence embedding delivery.