Pavithra Krishnamurthy

About

Built and shipped customer-facing AI systems in banking, including an AI fraud support workflow and virtual assistant at Citi/Citibank that handled over 1 million conversations per month. Combines full-stack engineering, LLM/RAG architecture, and strong compliance-minded design with PII masking, audit logging, confidence thresholds, and human escalation for regulated environments.

Key Strengths

• End-to-end ownership of AI-powered fraud alert workflows
• Full-stack development across React/Angular frontend and Node.js/Spring Boot backend
• Strong TypeScript API and interface design for multi-team integrations
• Production deployment of customer-facing AI assistants in banking
• LLM integration with Azure OpenAI for intent understanding and contextual responses
• Banking compliance focus with PII masking, audit logging, and human escalation paths
• RAG architecture for enterprise document search and question answering
• Production AI monitoring using accuracy, escalation rate, latency, and user feedback
• Iterative prompt and retrieval tuning based on conversation log analysis
• Pragmatic AI automation boundary-setting for high-risk vs low-risk workflows
• Cross-functional alignment with product, operations, and compliance teams in ambiguous environments
• Scaled AI assistant handling over 1 million conversations per month
• Real-time fraud notifications delivered in under 2 seconds
• Improved first contact resolution and reduced average handling time
• Owned AI-powered full-stack customer support and fraud platform in production at Citibank
• Built across frontend, backend, and infrastructure on the same product
• Strong enterprise AI orchestration and human-escalation workflow design
• Real-time distributed systems architecture using Kafka event streaming
• Scaled production platform to 1M+ monthly customer interactions
• Reduced average handling time by about 38%
• Improved first contact resolution and reduced manual workload
• Designed intuitive dashboards for non-technical operations and fraud users
• Balanced AI response quality against latency with hybrid rules plus LLM approach
• Improved maintainability by moving frontend complexity into orchestration APIs
• Strong observability mindset with Prometheus, Grafana, logging, and alerting
• Experience shipping in ambiguous zero-to-one enterprise environments
• Cross-functional collaboration with product, operations, compliance, and business stakeholders
• Performance optimization with Redis caching and PostgreSQL query tuning

Experience

Education

Skills

Languages

Certifications