Pavan Kalyan
Ai Data Scientist @State Street
Signup · Get unlimited contacts
WORK HISTORY
Ai Data Scientist @State Street
Built a counterparty credit risk scoring model using XGBoost on enriched trade exposure data (PD, MTM, collateral). Integrated via FastAPI into a front-office dashboard, reducing analyst review time on high-risk flags by ~35%.Developed a TF-IDF + gradient boosting pipeline to classify SWIFT messages, trade confirms, and settlement instructions by type and urgency. Automated routing handled 70%+ of daily volume within the first month.Implemented a LangChain-based RAG system with Pinecone for querying regulatory and policy documents. Enabled natural language search with source attribution, reducing lookup time from hours to under 2 minutes.Built a client attrition early-warning model using Random Forest and XGBoost with SHAP validation. Achieved 3.8x top-decile lift vs baseline on held-out data before production deployment.Automated compliance document review using LangGraph with multi-step agent workflows (parsing, clause extraction, policy matching, structured output). Reduced turnaround time from 2 days to <4 hours.Designed an ETL validation framework (FK checks, duplicates, outliers), reducing data-related training failures by ~30% and eliminating recurring pre-production issues.Enhanced transaction monitoring using Isolation Forest and DBSCAN to detect anomalies in settlement activity, integrated into existing compliance workflows without disrupting throughput.Replaced manual Excel-based portfolio reporting with a Python/Pandas pipeline, cutting generation time from ~2 days to 93%.Used Cursor AI and GitHub Copilot for FastAPI scaffolding, test generation, and prototyping, with strict review for governance and security compliance.
EDUCATION
University of North Texas
Master's degree, Computer Science
ABOUT PAVAN KALYAN
AI Data Scientist with 5 plus years of experience across financial services, working at State Street Corporation and Hexaware Technologies on end-to-end machine learning pipelines, NLP-driven automation, predictive modeling, and AI-augmented analytics.Hands-on across the full model lifecycle from data validation and feature engineering through production deployment and drift monitoring, with strong exposure to LLM integration and RAG-based solutions that extend traditional data science workflows.Comfortable working across cross-functional teams spanning data engineering, risk, compliance, and product to deliver models that are technically sound and genuinely useful to the business. Actively leverages AI-assisted development tools to move faster without cutting corners on quality.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.