Revathi Thangadurai

Architected and deployed an agentic AI system on AWS using OpenAI GPT-4 and LangChain to automate complex client request handling across planner, researcher, and writer agents. Reduced average handling time for client research requests by 50% and implemented FAISS/Weaviate/PGVector RAG over internal policy and regulatory documents with monitoring and compliance guardrails.

Architected an intelligent document automation system for an insurance carrier, generating personalized policy summaries and renewal/claim updates using React and FastAPI. Built NLP pipelines for intent classification and inquiry routing, established end-to-end MLOps (GitHub Actions CI/CD), and implemented monitoring and cost/quality tracking for NLP and content generation components.

Designed and deployed production-scale computer vision pipelines using OpenCV and PyTorch, including fine-tuning ResNet50 for image-to-text generation. Built secure RESTful services with RBAC and JWT, integrated Google Vision API with AWS Rekognition fallback for robust inference, and operationalized batch inference with MLOps on AWS (EC2, Lambda, S3, RDS) with monitoring and Docker-based containe

Built real-time streaming data pipelines using Apache Kafka and implemented scalable ETL with PySpark for large-scale transformation and analytics. Developed FastAPI REST services for data access and automated BI reporting through scheduled jobs, reducing manual report generation effort by ~30%, while adding monitoring/alerting and Docker-based deployments.

Created interactive KPI dashboards and reports using Tableau and Google Looker Studio, driven by SQL extracts and Python-based exploratory analysis (Pandas/NumPy). Delivered end-to-end analytical workflows (cleaning, validation, statistical analysis) and provided actionable recommendations to improve business efficiency while maintaining version control for analytical code and queries.