Nivas Mankali

AI/ML Engineer with 3+ years of experience developing and deploying machine learning, NLP, and Generative AI solutions across healthcare and financial services. Experienced in Python, PyTorch, TensorFlow, Scikit-learn, Hugging Face Transformers, ClinicalBERT, LLMs, RAG, and LangChain, with hands-on expertise in document intelligence, fraud detection, and predictive modeling. Skilled in building scalable AI/ML pipelines using PySpark, FastAPI, Docker, MLflow, and Apache Airflow across Azure and AWS environments

AI/ML Engineer - United Health Group - IN

(2024-10)

• Developed a medical document intelligence pipeline using Python, Pandas, and NumPy to preprocess clinical notes, authorization documents, and physician reports for AI-driven analysis
• Fine-tuned Hugging Face Transformers and ClinicalBERT models to extract diagnoses, procedures,medications, and clinical entities from unstructured healthcare documents
• Built a Retrieval-Augmented Generation (RAG) solution using LangChain and FAISS for semantic document search and context-aware information retrieval
• Implemented an AI-powered documentsummarization system using Azure OpenAI Service and prompt engineering, reducing manual review time for clinical analysts
• Deployed scalable LLM inference APIs using FastAPI and Docker, enabling secure integration with internal healthcare applications
• Automated document ingestion and inference workflows using Apache Airflow, improving pipeline reliability and processing efficiency across high-volume datasets.
• Optimized large-scale NLP processing with PySpark, accelerating document transformation and feature generationfor enterprise AIworkloads
• Managed model versioning, experiment tracking, and performance monitoring using MLflow, ensuring reproducible deployments and consistent model governance

ML Engineer - Deloitte - India

(2022-01 - 2023-11)

• Built data preprocessing and feature engineering pipelines using Python, NumPy, Pandas, and SQL to clean and prepare over 40 million credit card transaction records, creating high-quality datasets for fraud detection model development
• Trained and evaluated machine learning models using Scikit-learn, TensorFlow and XGBoost, comparing multiple algorithms and tuning hyperparameters to improve fraud detection precision by 12% while reducing false positives
• Processed large-scale transaction data using PySpark to perform feature generation and distributed data transformations, reducing data preparation time by 30% and improving model training efficiency
• Developed and deployed fraud prediction services using FastAPI, Docker, and AWS SageMaker, enabling real-time risk scoring with average inference latency of under 200 ms for production workloads
• Managed machine learning experiments and model lifecycle activities using MLflow and Git, tracking 100+ model training runs and maintaining reproducible workflows for collaboration with senior ML engineers
• Monitored model performance using Amazon CloudWatch and collaborated with fraud analysts and developers through JIRA to investigate prediction issues and implement model improvements, contributing to an estimated 8% reduction in fraud-related financial losses over successive model releases

Master - Information Systems - Indiana Institute of Technology University

Bachelor of Technology - Information Technology - JNTUH