Tarak Sai Varigonda
Ai Engineer @Mastercard
Signup · Get unlimited contacts
WORK HISTORY
Ai Engineer @Mastercard
US
Built real-time card-transaction risk models using PyTorch and CatBoost, processing 30M monthly events and improving chargeback detection by 14% while maintaining sub-120 ms authorization latency.Engineered a scalable feature engineering pipeline on Kafka and Flink (8K events/second), reducing model data freshness lag from 6 hours to 20 minutes for fraud scoring.Deployed automated model retraining workflows with Airflow and Kubernetes with weekly drift checks, improving approval-rate lift by 9% without increasing customer friction.Developed a GenAI merchant-support assistant using Azure OpenAI and LangGraph, cutting average case resolution time by 22% for dispute inquiries.Built a retrieval-augmented knowledge system with Milvus and sentence-transformer embeddings to enable compliance-safe answers for card regulations and reduce manual review effort.Optimized prompt orchestration and evaluation pipelines for internal analytics copilots, improving response accuracy and lowering hallucination rates.
EDUCATION
Cleveland State University
Master of Science - MS, Computer Science
CMR Institute Of Technology
Bachelor of Engineering, Electrical, Electronics and Communications Engineering
ABOUT TARAK SAI VARIGONDA
AI Engineer with experience building production-grade machine learning systems for real-time risk, fraud, and decisioning. I take models from idea to deployment—designing feature pipelines, training and validating models, and operationalizing them with monitoring, retraining, and governance so they perform reliably in high-throughput environments.I work on low-latency risk models and scalable data pipelines, using Kafka and Flink for streaming, and PyTorch plus gradient-boosting methods (CatBoost/XGBoost) for modeling. On the MLOps side, I build automated workflows with Airflow, Kubernetes, Docker, and MLflow, including drift checks and performance tracking to keep models accurate as data evolves.I also build GenAI solutions that improve productivity and support workflows. This includes retrieval-augmented generation (RAG) systems using embeddings and vector databases (Milvus/FAISS), orchestrating multi-step logic with LangGraph, and integrating Azure OpenAI to deliver reliable, compliance-aware assistants. I focus on reducing hallucinations, improving answer quality, and measuring impact with clear evaluation.Tech: Python, SQL, PyTorch, CatBoost/XGBoost, Kafka/Flink, Spark, FastAPI, Docker, Kubernetes, Airflow, MLflow, RAG, LangGraph, Azure OpenAI.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.