Pavana Sree Prathyusha Kollu

About Pavana Sree Prathyusha Kollu

AI/ML Engineer with experience building and deploying machine learning and AI solutions across banking, payments, public sector, retail, aviation, and consulting. I specialise in taking models from concept to production and delivering reliable, scalable AI systems in high-volume, regulated environments.My work spans Generative AI, Large Language Models (LLMs), NLP, and Retrieval-Augmented Generation (RAG). I’ve built LLM-powered applications such as chatbots, document intelligence workflows, and vector search systems using LangChain, embeddings, and vector databases like pgvector and ChromaDB. I focus on practical, production-ready GenAI solutions with strong performance, governance, and reliability.I have a strong foundation in machine learning, deep learning, and data science, with hands-on experience in fraud detection, transaction risk, credit risk, AML, anomaly detection, forecasting, and predictive analytics. I work extensively with Python, SQL, PySpark, Scikit-learn, XGBoost, TensorFlow, and Keras, applying feature engineering, model tuning, and evaluation to build robust models at scale.I design end-to-end ML pipelines covering data ingestion, training, deployment, and monitoring for both real-time and batch systems. I’ve built low-latency inference services using Python, FastAPI, and REST APIs, and deployed them as cloud-native, containerised services.My cloud experience includes AWS and Azure, working with services such as S3, EC2, Lambda, SageMaker, Bedrock, EMR, Redshift, CloudWatch, and Azure AI / Azure ML. I’m experienced in MLOps and LLMOps, including CI/CD, model versioning, monitoring, drift detection, and automated deployments using Docker, Kubernetes, GitHub Actions, and Jenkins.I enjoy working closely with product, risk, and engineering teams to translate business and regulatory needs into clear technical designs and well-documented, auditable AI systems. I’m driven by building AI solutions that are not only accurate, but also scalable, explainable, and production-ready.

Experience

1. Sr Ai Ml Engineer

   Fiserv

   Dec 2023 — Present · Milwaukee, WI, US

   Designed and deployed an Intelligent Payment Routing & Authorization Optimization platform using Python, Scikit-learn, PySpark, and AWS, improving authorization success rates by 14% and reducing network decline losses across high-volume payment transactions.• Architected a cloud-native, microservices-based ML platform with real-time inference APIs, batch training pipelines, and event-driven ingestion to enable context-aware, adaptive decisioning across distributed payment authorization systems.• Built scalable data ingestion and feature engineering pipelines integrating payment APIs, transaction event streams, Amazon S3, SQL, and NoSQL databases, processing billions of records for real-time inference and historical analytics.• Trained, evaluated, and optimized Gradient Boosting, XGBoost, and Logistic Regression models, applying feature selection, hyperparameter tuning, and experimentation (A/B testing, shadow deployments) to balance approval uplift, latency, and financial risk.• Deployed and operated production ML services using Docker, Kubernetes (EKS/ECS/Fargate) with CI/CD, MLOps, monitoring, and model governance, ensuring low-latency, scalable, and auditable AI systems in regulated payment environments.

Education

• Gandhi Institute of Technology and Management (GITAM)

  Bachelor of Technology - BTech

  2017 — 2021

• University of New Haven

  Master's degree

  2024 — 2025