Shyam Lunagariya

About Shyam Lunagariya

A hardworking and passionate research engineer with strong organizational skills, Meticulous, highly organized, motivated, and diligent machine learning enthusiast with significant background in Data Science. Demonstrated success in identifying relationships and building solutions to real-world problems Obdurate in achieving goals Critical thinking practitioner in engineering SKILLS • Programming: Python, SQL, C++, C# • Libraries: Scikit-Learn, TensorFlow, Pandas, PyCaret, MPI4PY, DASK, NumPy, PyTesseract, SciPy, PyTorch, Tkinter, selenium • Database: MySQL, PostgreSQL • Visualization: Tableau, Matplotlib • Cloud: Agave Cluster (ASU Research Computing) • Others: Git, GitHub, VS Code

Experience

1. Software Development Engineer

   Amazon

   May 2022 — Present · Palo Alto, CA, US

   Model Assertion: Developed a critical feature to validate ML modules for machine/deep learning training jobs before submittingto SageMaker, proactively identifying and preventing errors before job execution.• Achieved a remarkable IMR cost reduction of $2 million in 2023 by eliminating failed training jobs due to configuration issues,resulting in a 60% reduction in overall job failures.• Fault Tolerant Training: Implemented checkpoint-based resume functionality using native PyTorch Lightning methods withinthe team’s custom codebase, saving a substantial amount in IMR costs by avoiding full model retraining.• Introduced a job categorization system distinguishing between retriable and non-retriable failures, enabling an intelligent job retrystrategy which reflected the IMR cost savings of $10 million in 2023 by minimizing retraining efforts.• Model Dashboard: Established the model dashboard as the central repository for all ML-related artifacts, incorporating MLflowand Model Registry for comprehensive tracking.• Designed and constructed a CloudWatch Dashboard using AWS CDK, offering real-time visibility into training job statusesand model performance metrics.• Developed custom Lambda scripts for monitoring training jobs across various stages and efficiently managed resource utilization,ensuring optimal usage of reserved instances.• Codebase Auto-Formatting: Implemented and configured a Git hook to enforce code formatting standards, integrating iSortand Black for consistent and clean code commits

Education

• Arizona State University

  Master of Science - MS, 3.79

  2019 — 2021

• National Institute of Technology Surat

  Bachelor of Technology (B.Tech.), Mechanical Engineering, 7.6

  2014 — 2018