➥ Developed Language Model to provide insightful responses while implementing End-to-end processes from Data Curation, Model Training, and Instruction Finetuning to Evaluations. Achieved 79% accuracy in Healthcare Oral domain. For Data Curation, developed an ETL pipeline, used REST APIs, and conducted statistical analysis to curate the data. In the model development process, used PyTorch for implementation, Architecture of Transformer with RoPE embeddings, bitsandbytes framework for model quantization, Docker for containerization, and Deployed on SageMaker.

➥ Implemented Semantic Search to optimize accuracy and efficiently retrieve relevant research papers to speed up the research using Retrieval Augmente Generation (RAG). In the development process, converted to embeddings with SentenceTransformer from HuggingFace API, stored the vectors in ChromaDB in persistent Index, Documents retrieval using advanced filtering and metadata, Deployed on EC2 instance, and integrated S3 storage system to access VectorStore.

➥ Performed Cluster Modeling for a recommendation system that streamlines the research discovery process. Utilized Cluster Modeling and Statistical Analysis of dense regions in the data, visualized with Matplotlib. To analyze and improve performance, reduced the dimensionality using the Principal Component Analysis (PCA) technique and optimized the attributes and hyperparameters with memory optimization. Used the DBSCAN estimator from scikit-learn to obtain the best estimator. The modeling effectively suggests relevant papers to users based on dense regions in the data, making it a valuable tool for medical research.