Saurin Patel

Design and productionize scalable ML and AI systems across distributed cloud environments. • Designed and deployed enterprise ML and LLM-powered systems using Python, PyTorch, and foundation models (GPT-4, LLaMA), supporting customer-facing and internal analytics platforms. • Built scalable training, evaluation, and monitoring pipelines using MLflow, Airflow, Spark, and Databricks — reducing pipeline failures by 40% and improving reproducibility across teams. • Engineered Agentic AI and RAG architectures (LangChain, LangGraph, vector databases, tool-calling), improving response consistency and relevance by 30% in production workflows. • Developed structured model evaluation frameworks (offline + online) measuring performance, drift, bias, hallucination, and stability to ensure Responsible AI compliance. • Fine-tuned transformer-based models in PyTorch, optimizing inference latency and compute cost while maintaining model accuracy. • Built distributed data engineering pipelines (Spark, PySpark, Delta Lake, Microsoft Fabric) supporting batch and near real-time workloads with ACID-compliant ETL/ELT design. • Productionized containerized ML services (FastAPI, Docker, Kubernetes) integrated with Azure and AWS; owned services end-to-end from experimentation to deployment and monitoring. • Implemented automated regression detection and model health dashboards, reducing production incidents by 35%. • Partnered cross-functionally with ML engineers, platform teams, and business stakeholders to translate research models into reliable production systems.

Built a full-stack RAG system for financial document Q&A, integrating LangGraph agents, FastAPI backend, and Next.js frontend, increasing research efficiency by 60%. Designed an agentic pipeline with LangChain and LangGraph for multi-vector retrieval, sub-question decomposition, and external API integration using LLMs. Developed a scalable Financial Research Assistant leveraging Llama 3.2 Vision and GPT-4o to analyze SEC filings and analyst reports, reducing research time by 60% and improving accuracy by 25%. Deployed automated model training via Airflow and MLflow, enabling consistent MLOps workflows. Developed backend services and scalable APIs using modern software engineering practices. Built cloud-ready applications leveraging containerization and CI/CD pipelines. Contributed to AI-integrated features within distributed system architecture.

Designed and developed full-stack web applications using React.js, Node.js, and SQL, improving page load time by 40% and increasing engagement by 30%. Integrated Docker-based CI/CD pipelines, ensuring consistent deployment across environments. Developed automated Cypress regression testing frameworks, improving QA reliability and deployment velocity. Optimized database performance via advanced SQL query tuning and stored procedure refactoring, improving dashboard performance by 70%. Contributed to scalable backend services and API development. Tech: React.js, Node.js, SQL, Docker, CI/CD, Cypress

Led data preprocessing for machine learning models (CNNs, Random Forest, SVM), improving input data quality. Enhanced model performance by 25% through data cleaning, standardization, and augmentation. Conducted extensive data analysis using Python to extract meaningful insights and inform model development. Achieved 92.9% accuracy and reduced errors by 40% in Land Use and Cover Classification with region-based training. Presented research findings at team meetings, effectively communicating complex technical concepts to both technical and non-technical stakeholders. Actively participated in weekly knowledge-sharing sessions, staying abreast of the latest advancements in machine learning, remote sensing, and geospatial technologies.

Improved brain tumor identification accuracy in MRI images to 89.2% using advanced analysis techniques. Optimized YOLOv7 model for better performance using Python, TensorFlow, and OpenCV. Conducted exploratory data analysis (EDA) to identify key features and correlations in medical imaging data. Developed custom Python scripts to automate data preprocessing tasks, reducing manual effort by 30%.

Conducted hands-on testing, creating test cases for APIs and tracking bugs with Postman. Utilized strong communication skills for cross-functional collaboration. Led a team of four in project management, designing and building web components with React, and adopting Agile/Scrum methodologies.