Senior ML Engineer - Distributed AI & MLOps Expert

Pavani Billapati

*******************@*****.*** +1-502-***-**** linkedin.com/in/pavani-billapati-746604144/ SUMMARY

• Senior Machine Learning Engineer with 7+ years’ experience delivering AI/ML solutions at Google, NVIDIA, Deloitte, and Johnson & Johnson, specializing in distributed ML pipelines, deep learning (Transformers, GNNs, LSTMs), reinforcement learning, Retrieval-Augmented Generation (RAG), GPU acceleration (CUDA, RAPIDS), and MLOps automation using Python, PyTorch, Kubeflow, Vertex AI, and AWS SageMaker.

• Proven ability to optimize model performance and efficiency—including 55% faster training times, 35% higher accuracy in RAG systems, and $100K/year cost savings—across domains such as search relevance, multi-modal modeling, predictive analytics, and time-series forecasting, leveraging tools like XGBoost, LightGBM, Optuna, Ray Tune, and SQL on platforms including BigQuery, Firestore, MongoDB, PostgreSQL, and Redshift.

• Skilled in end-to-end product delivery, cloud platforms (GCP, AWS, Azure), CI/CD (Docker, Kubernetes, GitHub Actions, Jenkins), and cross-functional leadership, with astrong track record in team mentorship, workflow optimization, and aligning AI initiatives with business goals to launch high-adoption, ML-powered products. TECHNICAL SKILLS

Programming & Scripting Languages: Python (NumPy, Pandas, Matplotlib, SciPy, scikit-learn, TensorFlow, PyTorch, Keras, NLTK, Plotly, PyMC3), R, Scala, SQL (MySQL, PostgreSQL, Oracle, T-SQL, PL/SQL), NoSQL (MongoDB, Cassandra, DynamoDB). Machine Learning & Artificial Intelligence: Supervised Learning, Unsupervised Learning, Deep Learning (CNNs, RNNs, LSTMs, Transformers, GNNs), Large Language Models (LLMs), Natural Language Processing (NLP), Time Series Analysis, Feature Engineering, Generative Models, Reinforcement Learning, Retrieval-Augmented Generation (RAG), Anomaly Detection, Model Evaluation (Cross-Validation, Ensemble Methods, Hyperparameter Tuning), Prompt Engineering. Big Data & Data Engineering: Apache Spark, Hadoop, Hive, Kafka, Flink, Airflow, MapReduce, EMR, Sqoop, HBase, Talend, SSIS, AWS Glue, Kinesis, Informatica, DataStage, Prefect. Cloud Platforms & MLOps: GCP (Vertex AI, BigQuery, Firestore, Cloud TPUs), AWS (EC2, S3, Redshift, SageMaker, Lambda), Azure, Snowflake, Teradata, Kubeflow, MLflow, CI/CD Pipelines. Deployment, DevOps & Infrastructure: Docker, Kubernetes, Jenkins, GitHub Actions, Terraform, Ansible, Canary Deployments. Data Visualization & Analytics: Tableau, Power BI, Microsoft Excel, Plotly, NVIDIA Omniverse, SSRS, Exploratory Data Analysis

(EDA), A/B Testing, Predictive Analytics, Statistical Analysis. Tools, Frameworks & Libraries: Jupyter Notebook, RStudio, BERT, spaCy, TensorRT, Ray Tune, Optuna, LangChain, Gemini APIs. PROFESSIONAL EXPERIENCE

Google CA Senior Machine Learning Engineer June 2024 – Present

• Designed and deployed distributed ML training pipelines using Python (PyTorch, Scikit-learn), TensorFlow Extended (TFX), and Google Cloud TPUs, achieving 55% faster training times for key products including Google Search and Google Ads.

• Developed Transformer-based and Graph Neural Network (GNN) architectures for multi-modal search ranking, integrating text, image, and user behavior data; improved search relevance metrics by 22%.

• Built end-to-end MLOps pipelines with Vertex AI, Kubeflow, and custom CI/CD tooling, automating model monitoring, retraining, and canary deployments; reduced incident response time by 60%.

• Implemented Retrieval-Augmented Generation (RAG) systems using Gemini APIs and LangChain, combining proprietary search indices with LLMs; improved answer accuracy by 35% while reducing hallucinations.

• Engineered scalable data pipelines using BigQuery (SQL) and Firestore (NoSQL), increasing data retrieval speed by 50% for ML training and real-time inference workloads.

• Led mentorship programs for 5+ juniorengineers and cross-functional teammembers on ML bestpractices (PyTorch, TFX, Vertex AI), boosting team productivity by 50%.

• Collaborated with product managers, research scientists, and SRE teams to align AI initiatives with business objectives; launched 3 ML-powered features with 20% higher user adoption than projected. NVIDIA CA Machine Learning Engineer August 2023 – June 2024

• Built scalable machine learning pipelines using AWS SageMaker, EC2, S3, and NVIDIA GPU Cloud, improving data processing efficiency by 45% and enabling low-latency real-time model deployments for high-throughput applications.

• Developed GPU-accelerated reinforcement learning algorithms with NVIDIA Isaac Gym and PyTorch, achieving 2 faster training than CPU-based methods and 92% success rates in simulation-to-real transfer learning.

• Engineered parallel training and inference algorithms in CUDA C++, optimizing GPU kernels to reduce latency by 20% in large- scale recommendation systems.

• Accelerated data pipelines using CUDA/RAPIDS, cutting training times by 15%, and implemented real-time feature stores with MongoDB and NVIDIA Omniverse to enhance ML visualization and dynamic decision-making by 8%.

• Optimized model performance with XGBoost and LightGBM using distributed training and hyperparameter tuning tools

(Optuna, Ray Tune), improving accuracy by 12% for mission-critical workloads.

• Deployed ML models at scale using Docker, Kubernetes, and CI/CD workflows (GitHub Actions, Jenkins), automating release cycles and reducing deployment time by 30% while maintaining MLOps best practices. Deloitte India Machine Learning Engineer May 2020 – July 2022

• Developed time-series forecasting models for ERP and inventory optimization, using predictive analytics and real-time demand forecasting to reduce inventory holding costs by 12% for enterprise clients.

• Improved ML model accuracy by 18% and reduced training time by 30% across 8+ client projects through hyperparameter tuning (Bayesian Optimization, Grid Search) and feature selection techniques.

• Built MLOps workflows using Kubeflow and MLflow for automated model deployment, monitoring, and pipeline stability, reducing model downtime by 25%.

• Conducted A/B testing on NLP-driven customerengagement models leveraging Azure Cognitive Services, increasing customer satisfaction by 15%.

• Implemented anomaly detection for IoT sensor data using Apache Spark and Hadoop, reducing equipment downtime by 10% in large-scale industrial operations.

• Automated ML workflows with Apache Airflow, designing DAGs for data ingestion, feature engineering,and retraining, reducing manual intervention by 60%.

Johnson & Johnson India Data Scientist Nov 2017 – May 2020

• Contributed to a development of NLP-based sentiment analysis models using Word2Vec and LSTM networks to analyze patient feedback from clinical trials and social media, improving insight extraction efficiency by 30% for pharmaceutical teams.

• Built end-to-end predictive modeling pipelines with Python (Scikit-learn, TensorFlow), automating drug efficacy prediction and reducing manual validation time by 25%.

• Enhanced supply chain forecasting with ARIMA and Prophet models, reducing inventory costs by 18% ($100K/year) and lowering stockouts by 12% in India’s pharmaceutical distribution network.

• Implemented anomaly detection using PCA and clustering algorithms for manufacturing equipment monitoring, decreasing unplanned downtime by 20% via predictive maintenance.

• Designed and optimized complex SQLqueries (window functions, CTEs) in PostgreSQL and Redshift to process clinical trial data, reducing data preparation time for research teams by 40%.

• Collaborated with pharmaceutical researchers, supply chain managers, and IT teams to transform data insights into actionable business strategies, increasing stakeholder adoption of analytics solutions. EDUCATION

Master’s in Data Science University of New Haven, Data Science, CT, USA. CERTIFICATIONS

• Google Cloud Certified - Associate Cloud Engineer

Contact this candidate