Lavanya Pobbathi

Senior Data Scientist (through Infosys) @Citi

Irving, TX, US

Architected and shipped an NLP engine for RFQ (Request-for-Quote) chat to Trade Object conversion. Built a hierarchical RFQ/product classifier, DeBERTa-based NER per product, entity normalization, ref-data enrichment (CVD/SMC), and object construction. Covered 60+ products across Bonds, FX, Equity Derivatives, and SBL, exposed via REST APIs for analytics and straight-through processing.• Engineered production scale and model excellence. S3-hosted model artifacts, load-balanced V100 GPU servers (Prod/COB), CI/CD with versioned models, telemetry and drift checks. Improved macro-F1 to ~0.85 (from 0.56 baseline) across 25 entities—boosting extraction accuracy and enabling higher STP rates for trading operations.• Developed an SDLC automation framework using LangChain, LlamaIndex, LangFlow and Azure OpenAI integration by using LLM models like GPT-4.5, Llama 3.2, Claude 3.1 Sonnet API key and created AI agents, streamlining requirements gathering, user story generation,and test case creation, reducing documentation processing time by 40% and improving end-to-end traceability.• Designed & deployed an intelligent Retrieval-Augmented Generation (RAG) and RAGAS frameworks with advanced flowchart processing capabilities, enabling accurate interpretation of complex SDLC documentation and improving process flow comprehension by 35% for development teams.

University of North Texas

Masters