GenAI Based SLR Tool: Pharma-NLPRole: ML LeadLed a multidisciplinary team to develop ‘Pharma-NLP,’ an AI-driven Systematic Literature Review (SLR) tool that enhances the literature review process for medical writers.Created an AI copilot for article screening, data extraction, summarization, and report generation, reducing SLR task time by 40%.Integrated diverse data types to support clinical trials, case reports, and meta-analyses, ensuring comprehensive reviews.Collaborated with customer SMEs, clinical analysts, developers, and DevOps teams for seamless integration and continuous improvement.Implemented GenAI technologies to enhance platform capabilities and keep pace with the latest advancements.Defined the ML development strategy and data accuracy scoring, ensuring reliable outputs.Increased customer satisfaction by adding new features, boosting user engagement by 25%.Fitment Predictor for Teeth AlignersRole: ML LeadLed a team in developing a deep learning-based solution using YOLO and transfer learning to predict teeth aligner fitment.Established ML strategy and data annotation processes for high precision in model training.Collaborated with SMEs and validation teams to improve model accuracy by 30% through rigorous validation.Worked with DevOps to define MLOps requirements, reducing deployment time by 50%.Data Extraction from Unstructured DataRole: ML LeadLed development of a GenAI and LLM-based application for extracting data from medical bills using Claude3 and AWS Textract.Formulated the ML strategy and accuracy scoring, achieving a 35% improvement in data extraction accuracy.Worked with SMEs and validation teams to ensure data reliability and accuracy.These roles demonstrate my ability to lead ML projects, collaborate effectively, and implement innovative solutions that improve efficiency and accuracy.

NLP-aided Abstract Screening for SLRsRole: ML EngineerPioneered the development of an NLP technique to automate and expedite Systematic Literature Reviews (SLRs), facilitating live literature review.Designed and trained a binary text classifier model using pretrained Transformer-based models to screen abstracts of relevant studies.Optimized training and inference pipelines, significantly enhancing model efficiency and reducing processing time.Deployed the solution as a REST API and containerized the application, ensuring seamless integration and scalability.Information Extraction for Biomedical ArticlesRole: ML EngineerLed the annotation process for Named Entity Recognition (NER) training data, ensuring high-quality labeled data.Experimented with Spacy and transformer-based models to enhance performance, achieving superior accuracy in extracting 14 different entities.Developed custom NER models and deployed them as REST APIs, enabling efficient information extraction from biomedical texts.Tissue Segmentation in Arthroscopy Videos Role: ML EngineerCollaborated with a leading consulting firm to develop Computer Vision tools for subject matter experts.Performed data preprocessing and labeling, ensuring high-quality training data for the model.Trained an image segmentation model using Detectron2 through transfer learning, successfully segmenting ACL tissues in arthroscopy videos.Steel Tube Surface Defect ClassifierRole: ML EngineerDeveloped an image classifier using labeled surface defect images, utilizing the TensorFlow framework.Optimized the model for edge computing devices using TensorRT, ensuring high performance and efficiency.