Gayathri Sukuri

Built agentic AI and multimodal LLM solutions across transaction, merchant, and customer datasets, delivering real-time payment intelligence and revenue forecasting. Developed predictive models using TabPFN, CatBoost, LightGBM, and Temporal Fusion Transformers to improve merchant lifecycle prediction accuracy by 34%. Engineered real-time streaming pipelines with Databricks, Apache Iceberg, Delta Live Tables, Apache Flink, Kafka, Snowflake, AWS Bedrock, and Ray to support low-latency inference. Implemented AI copilots with LangGraph, LlamaIndex, vector databases, MCP, semantic caching, hybrid search, and OpenAI-compatible APIs, reducing analytical response times by 29%. Optimized production ML with Bayesian optimization, MLflow, Arize Phoenix, SHAP, and continuous monitoring, reaching 0.94 ROC-AUC and 96% precision. Established scalable MLOps workflows using GitHub Actions, monitoring, automated testing, and governance controls to improve deployment reliability.

Architected end-to-end ML pipelines using Python, SQL, feature engineering, and MLOps, deploying predictive models through FastAPI with 92% accuracy and reducing inference latency by 35%. Engineered real-time Kafka streaming pipelines processing 5M+ daily events, optimizing distributed ingestion and feature delivery by 40% to enable low-latency ML inference and analytics dashboards. Automated Docker-based CI/CD with GitHub Actions, reducing deployment cycles from two weeks to four hours (75% release velocity improvement) and building analytics APIs to increase actionable ML insights across production applications.

Engineered scalable SQL/Python/ADF/Databricks ETL pipelines processing terabyte-scale datasets across 50+ sources to support ML model development and production deployment. Developed interactive Power BI and Tableau dashboards for churn, revenue, SLA compliance, and operational KPIs, automating reporting and supporting data-driven executive decision-making. Applied statistical analysis, regression modeling, anomaly detection, and Spark-based big data processing to identify patterns, optimize queries threefold, and uncover $2M savings. Conducted exploratory data analysis, feature engineering, data validation, and model performance analysis using Python, SQL, and Spark to improve dataset reliability for scalable ML and analytics.