Zarana Nakrani

Architected a production-grade RAG platform using FastAPI, LangChain, LlamaIndex, Pinecone, and OpenAI APIs, reducing knowledge retrieval latency by 68% and improving response relevance scores by 42% across enterprise support workflows. Engineered scalable AI microservices with Python, Docker, AWS EKS, Redis, and PostgreSQL to support 5M+ monthly API requests and increase throughput by 55%. Developed agentic AI workflows using LLM orchestration, vector search, prompt engineering, and Bedrock integrations, automating 78% of manual knowledge-management tasks and saving 1,200+ operational hours annually. Optimized real-time data pipelines using Kafka, Airflow, and ETL frameworks, accelerating model feature availability by 60% and improving prediction freshness. Led deployment automation with Jenkins CI/CD, Docker, AWS Lambda, and Infrastructure-as-Code, reducing release cycles from days to under 2 hours while achieving 99.95% availability. Improved ML performance via feature engineering,

Developed backend microservices in Python/FastAPI and Node.js/Express.js, improving API throughput and reducing transaction processing time from minutes to seconds for 120k+ monthly transactions. Optimized distributed event-driven pipelines leveraging Apache Kafka and Redis caching, improving real-time data processing throughput by 42% while lowering API latency by 30%. Deployed containerized services on AWS EKS with Docker and Jenkins-driven CI/CD automation, accelerating release cycles by 50% and cutting deployment failures by 60%. Built responsive internal dashboards using React and backend APIs for real-time incident response visibility and improving operational SLA adherence by 25%. Implemented observability and monitoring using centralized logging, metrics, and alerting tools. Developed and maintained Foundry data pipelines and contributed to application development within the platform, leveraging Python-based transforms and SQL logic to deliver clean, analysis-ready datasets.

Built end-to-end predictive analytics solutions using Python, Scikit-Learn, Pandas, and PostgreSQL, improving forecasting accuracy by 25% and enabling data-driven decisions across business units. Designed RESTful APIs with FastAPI and Node.js, including JWT authentication and RBAC, reducing unauthorized access incidents by 90%. Implemented MQTT-based IoT ingestion and Kafka streaming to process 10M+ daily events, decreasing data delivery delays by 72% for operational monitoring systems. Automated large-scale ETL workflows with Airflow, Redis caching, and PostgreSQL optimization, reducing batch-processing times by 58% and increasing reliability to 99.8%. Integrated ML models into customer-facing applications using TensorFlow and Flask/FastAPI services and cloud infrastructure, improving recommendation effectiveness by 33% and user engagement by 21%. Modernized full-stack platforms with React.js, TypeScript, Tailwind CSS, and Express.js, improving responsiveness by 45% and reducing load