Rahul Ponnaluri

• Design and develop Python-based backend services that support AI-driven audit and risk analysis workflows, focusing on reliability, traceability, and clean separation of business logic and inference logic. • Build and maintain data ingestion pipelines in Python to process structured historical audit data (CSV and tabular formats), ensuring consistent schema handling and data quality validation. • Integrate Azure services with Databricks to enable scalable access, preprocessing, and analysis of large historical datasets used for similarity-based risk assessment. • Implement similarity search logic to compare new audit observations against historical records, applying strict score thresholds to ensure only genuinely relevant past cases are considered. • Develop GenAI-assisted workflows using Large Language Models (LLMs) to analyze historical evidence and support risk-level classification for new audit observations. • Perform Prompt engineering across multiple LLM configurations to produce concise, explainable, and audit-friendly outputs without hallucination or unsupported assumptions. • Design backend decision logic that combines similarity scores and structured historical attributes, ensuring deterministic behavior before invoking LLM-based reasoning. • Build secure and scalable API endpoints using Python (FastAPI-style architecture) to expose AI-driven insights and risk summaries to downstream systems. • Integrate LangChain with backend services to manage prompt templates, token limits, and model configuration dynamically. • Design multi-step AI workflows using LangGraph to enable stateful execution of complex reasoning flows. Use LangGraph to define deterministic execution paths, preventing uncontrolled LLM behavior.

• Designed and deployed scalable machine learning pipelines using Azure Machine Learning (Azure ML) and Azure Data Factory, enabling automated model retraining and deployment across multiple banking applications. • Built and optimized deep learning models using PyTorch and TensorFlow, enhancing fraud detection accuracy and reducing false positives in high-volume transaction systems. • Integrated Azure Cognitive Services for real-time NLP-based chatbot solutions, improving customer service response time by 40% through intelligent automation. • Led the migration of legacy ML models to Azure Kubernetes Service (AKS) with MLflow integration, ensuring better scalability, version control, and resource optimization. • Developed real-time inference APIs using FastAPI and deployed them using Azure App Services, accelerating model response time for loan underwriting workflows. • Leveraged Azure Synapse Analytics and Azure Data Lake Storage Gen2 to source, transform, and curate massive datasets for machine learning feature engineering and experimentation. • Created advanced feature stores using Feast on Azure Databricks, standardizing and reusing features across ML models to increase productivity and consistency. • Built and monitored model drift detection frameworks with Evidently AI and Azure Monitor, allowing proactive retraining and improving model accuracy over time. • Implemented Responsible AI principles with Fairlearn, InterpretML, and Azure Responsible AI Dashboard, ensuring fairness, transparency, and compliance in ML models used in credit risk assessments. • Collaborated closely with data engineers, DevOps teams, and product owners in agile sprints to integrate AI models within cloud-native architectures using Azure DevOps (CI/CD). • Designed AutoML experiments using Azure AutoML to rapidly prototype and evaluate models for customer segmentation, increasing marketing campaign performance.

• Developed and deployed predictive models using Azure Machine Learning Studio to forecast patient readmission risks, leading to a 20% reduction in re-hospitalization rates. • Built robust data pipelines using Azure Data Factory and integrated them with Databricks for preprocessing structured and unstructured clinical data. • Leveraged Azure Cognitive Services and Text Analytics API to extract insights from patient feedback, claims notes, and medical documents for sentiment and intent analysis. • Designed and implemented deep learning models using TensorFlow and PyTorch for disease classification, improving diagnostic accuracy for chronic conditions by 15%. • Orchestrated end-to-end MLOps pipelines using Azure DevOps, enabling automated training, testing, and deployment of machine learning models across multiple environments. • Conducted model drift analysis and performance monitoring using Azure ML’s monitoring features to ensure high-quality, explainable models in production. • Engineered real-time anomaly detection systems using Azure Stream Analytics to flag irregular patient health patterns from IoT and wearable devices. • Applied AutoML capabilities in Azure to optimize model selection and hyperparameter tuning, significantly accelerating the experimentation lifecycle. • Built interactive dashboards in Power BI by integrating outputs from ML models, enabling business stakeholders to gain real-time insights on population health metrics.

• Developed and deployed predictive models using Python, scikit-learn, and XGBoost to enhance credit risk assessments, leading to a 15% improvement in default prediction accuracy. • Built machine learning pipelines on AWS SageMaker, enabling faster model training and deployment across customer segmentation use cases. • Utilized AWS Redshift, S3, and Glue for building a robust data ingestion and analytics environment supporting real-time and batch data sources. • Created advanced time series forecasting models for delinquency and prepayment prediction, using ARIMA, Prophet, and LSTM-based deep learning techniques. • Led the design of an internal fraud detection system leveraging unsupervised learning techniques such as Isolation Forests and Autoencoders to detect anomalies across millions of transactions. • Integrated Natural Language Processing (NLP) models using spaCy and BERT for analyzing customer feedback and automating sentiment classification. • Conducted A/B testing and experimental design to evaluate marketing strategies, resulting in optimized campaign targeting and improved ROI by 18%. • Built and maintained ETL workflows using AWS Glue and PySpark, ensuring reliable and scalable data preparation for downstream analytics. • Collaborated with Data Engineers and DevOps to implement CI/CD for ML models using AWS CodePipeline, improving model deployment lifecycle efficiency. • Applied feature engineering and dimensionality reduction techniques (e.g., PCA, t-SNE) to optimize model input spaces for high-dimensional financial datasets. • Designed customer lifetime value (CLV) models to guide personalized product recommendations and pricing strategies across lending portfolios. • Deployed models as RESTful APIs using Flask and hosted them on AWS Lambda with API Gateway, ensuring scalable, real-time scoring capabilities.

• Developed and deployed predictive models to optimize inventory management and product assortment, using Python, scikit-learn, and AWS SageMaker, improving demand forecasting accuracy by 20%. • Conducted customer segmentation analysis using k-means clustering, PCA, and hierarchical clustering to personalize marketing strategies, increasing email conversion rates by 15%. • Built ETL pipelines using AWS Glue and Lambda to ingest and process structured and unstructured retail data from POS, web, and mobile channels. • Applied time series forecasting techniques (ARIMA, Prophet) to forecast seasonal sales trends, enabling more informed procurement and logistics planning. • Designed and implemented A/B testing frameworks to assess promotional effectiveness across digital and in-store campaigns, leading to a 10% increase in ROI. • Utilized Amazon Redshift and Athena to query large-scale transactional data, enhancing insights into customer behavior and purchasing patterns. • Developed automated dashboards and visualizations using Tableau and Amazon QuickSight to track KPIs, product performance, and marketing impact in real time. • Created recommendation systems using collaborative filtering and content-based filtering methods to improve product discovery and upsell opportunities on Macy’s website. • Performed text mining and NLP on customer reviews and social media data using NLTK and spaCy to identify sentiment trends and product feedback. • Collaborated with business stakeholders to translate retail challenges into data science use cases, aligning technical deliverables with business impact. • Cleaned, wrangled, and analyzed large datasets using pandas, NumPy, and AWS Data Wrangler, reducing data preparation time by 30%.