• Designed and deployed an end-to-end GenAI-powered document intelligence platform on Google Cloud Platform to process large-scale Medicaid forms, leveraging Document AI, LLM-based entity extraction, and NLP pipelines for automated document understanding.

• Built LLM-driven entity extraction pipelines using Gemini foundational and pro models, applying prompt engineering, contextual parsing, and schema-based extraction to improve structured data retrieval from complex healthcare documents.

• Developed custom NLP models and document parsing workflows to handle multi-format Medicaid forms, improving extraction accuracy through iterative training, evaluation, and fine-tuning on production datasets.

• Implemented Retrieval-Augmented Generation (RAG) architecture to enhance document understanding by integrating structured knowledge sources with LLM outputs, improving response reliability and contextual accuracy.

• Architected scalable serverless AI pipelines using Cloud Run, Docker, and GCP services, enabling automatic scaling low-latency processing, and high availability for document processing workloads.

• Engineered dynamic request orchestration that intelligently switches between synchronous and asynchronous API processing based on document size and workload patterns, ensuring optimal throughput during peak traffic conditions.

• Built production-grade ETL and data processing pipelines handling 3000+ daily document processing events, integrating document ingestion, LLM inference, validation, and structured output generation.

• Implemented MLOps practices including model monitoring, performance evaluation, version control, and automated pipeline orchestration for reliable production deployments.

• Optimized system performance by tuning container concurrency, CPU/memory allocation, and request batching, improving pipeline efficiency and reducing operational costs.

• Collaborated with cross-functional teams to integrate AI-driven document processing workflows.

• Designed and developed deep learning–based recommendation systems using RNN, CNN, and Transformer architectures in TensorFlow and PyTorch to deliver personalized recommendations and improve content relevance.

• Built scalable data preprocessing and feature engineering pipelines using Python (Pandas, NumPy, Scikit-learn) to process large-scale structured and unstructured datasets for machine learning workflows.

• Implemented machine learning pipelines for recommendation and predictive analytics, leveraging ensemble learning techniques and gradient boosting models to enhance recommendation stability and model robustness.

• Applied advanced Natural Language Processing (NLP) techniques using BERT, Word2Vec, and transformer-based architectures for text classification, semantic analysis, and contextual understanding of customer interaction data.

• Integrated Large Language Models (LLMs) using OpenAI APIs and transformer-based models to enable intelligent document understanding, automated text generation, and contextual transaction analysis.

• Developed Retrieval-Augmented Generation (RAG) workflows combining semantic embeddings and contextual retrieval to enhance LLM responses and knowledge-based recommendations.

• Implemented model lifecycle management and experiment tracking using MLflow, enabling version control, reproducibility, and systematic evaluation of machine learning and LLM models.

• Designed scalable cloud-based ML infrastructure on AWS, integrating services such as Amazon S3, AWS Glue, Redshift, and Lambda to support distributed data processing and model deployment.

• Applied dimensionality reduction techniques including PCA and autoencoders to optimize high-dimensional feature spaces and improve training efficiency in deep learning pipelines.

• Built containerized ML deployment pipelines using Docker and Kubernetes, enabling scalable model serving, automated deployment, and efficient resource management in production environments.

• Developed machine learning and predictive analytics models using Python, SQL, Scikit-learn, Random Forest, XGBoost, and Gradient Boosting algorithms to support risk forecasting, churn prediction, and business intelligence across financial and operational datasets.

• Built end-to-end data science pipelines including data ingestion, preprocessing, feature engineering, model training, evaluation, and deployment for large-scale structured and semi-structured datasets.

• Conducted Exploratory Data Analysis (EDA) on large transactional and operational datasets using Python (Pandas, NumPy, Matplotlib, Seaborn) to identify patterns, anomalies, and key drivers influencing business outcomes.

• Designed and implemented deep learning models using TensorFlow and PyTorch, applying neural network architectures for predictive modeling, pattern recognition, and complex data relationships.

• Built Natural Language Processing (NLP) pipelines using Python, spaCy, NLTK, and transformer-based models for text classification, sentiment analysis, and intent detection across customer communications and compliance documents.

• Fine-tuned transformer-based Large Language Models (LLMs) including BERT and GPT architectures to extract contextual insights from financial documents and automate document analysis workflows.

• Implemented feature engineering and dimensionality reduction techniques to improve model generalization and handle high-dimensional data across machine learning pipelines.

• Designed and automated ETL workflows and data pipelines using SQL, Informatica PowerCenter, and Python, enabling reliable ingestion and transformation of enterprise data into centralized data platforms.

• Developed data visualization and analytics dashboards using Power BI, presenting key financial metrics, model insights, and business KPIs for decision-making and performance monitoring.

• Implemented model evaluation and monitoring frameworks using standard ML performance.