Rajkamal Kareddula

Automated ICD-10 code suggestions from electronic health records using multi-label classification and CAML architecture. Managed a team of 20 medical coders for NER data annotation using AWS SageMaker. Fine-tuned medical BERT models for NER tasks, achieving a 94.8% F1 score. Managed cloud infrastructure on GCP and AWS, and implemented Docker containerization.

Led binary classification projects for EHR data using Random Forest and TF-IDF. Developed a custom Python-based Rule-based Sentence Detection module. Won the SemEval 2023 LegalEval competition by pre-training and fine-tuning RoBERTa-based LLMs. Implemented RxNorm, SNOMED, and LOINC code suggestion systems using FastAPI.

Engineered a Status Detection Module for temporal relevance analysis in healthcare. Developed a Negation Detection Module that improved medical insurance claim accuracy by 40%. Created a rule-based Token Detection module for medical domain-specific tokens.

Implemented unsupervised clustering methods like Self-Organizing Maps and Mean Shift for Patient Case Similarity analysis. Researched techniques for improving word embedding quality and dimensionality reduction.

Utilized K-means clustering for topic detection in audio customer reviews. Transcribed audio using Google Speech-to-Text API and applied TF-IDF for cluster naming. Developed a web interface using Flask, HTML, CSS, and Bootstrap.