Rajesh Macherla

Developed an AI-powered insurance claims intelligence platform to process claim documents, extract key information, identify policy coverage, and provide decision-support insights for reviewers. Built GenAI and agentic AI solutions using Python, FastAPI, AWS Bedrock, LangChain, LangGraph, RAG, embeddings, and vector databases, including secure document ingestion pipelines with Amazon S3. Implemented OCR and NLP pipelines for extracting entities such as dates, claim amounts, policy numbers, and incident details. Designed production RAG pipelines grounded in policy clauses and enterprise knowledge using semantic retrieval with OpenSearch/vector search. Orchestrated multi-step agent workflows for document analysis, policy retrieval, claims validation, risk assessment, evidence verification, and summarization. Delivered REST APIs for RAG Q&A, agent execution, and asynchronous processing, while using PostgreSQL and vector storage to manage metadata, embeddings, and processing status. Added

Built a healthcare clinical document intelligence platform to automate processing of clinical notes, discharge summaries, lab reports, and medical records. Designed end-to-end ML/NLP workflows using Python, spaCy, Hugging Face Transformers, BERT, AWS services, FastAPI, PostgreSQL, Docker, and MLflow. Developed document classification models for automatic categorization, and implemented spaCy-based NER pipelines to extract diagnoses, medications, procedures, and clinical information. Created ingestion pipelines for structured and unstructured clinical documents, built text preprocessing for cleaning/normalization/tokenization, and applied labeling, feature engineering, training, and hyperparameter tuning. Implemented REST APIs for ingestion, classification, entity extraction, and real-time inference, storing outcomes in PostgreSQL and managing documents in Amazon S3. Containerized services, added CI/CD automation, and implemented monitoring/logging for API latency, processing failures,

Developed an intelligent financial risk and anomaly detection platform to identify unusual transactions, duplicate records, risk patterns, and high-risk financial activities for review. Built end-to-end ML solutions using Python, SQL, Pandas/NumPy, scikit-learn, XGBoost, Isolation Forest, PostgreSQL, AWS, Docker, and Power BI. Built SQL-based extraction and transformation pipelines, performed EDA to detect trends and anomalies, and engineered risk-focused features such as transaction frequency and historical deviation. Implemented anomaly detection with Isolation Forest and supervised classification models (Logistic Regression, Random Forest, XGBoost), evaluating with precision/recall/F1 and ROC-AUC. Produced risk scoring logic to categorize low/medium/high risk and added explainability using feature importance and SHAP. Designed PostgreSQL data models for storing engineered features, anomaly scores, and predictions, and delivered APIs to integrate outputs into downstream analytics. C