Conducted significant work on the development and enhancement of the microservices infrastructure for KLA's machine learning driven defect classification system, designed to scale for the training of 100K defects and infer upto 1M defects.
Enhanced data loading efficiency by 50% by employing asyncio framework in Python to parallelize the reader, which significantly decreased the training duration.
Developed a training engine featuring a unique user interface for setting model hyperparameters and enabling iterative training process.
Utilized Apache Airflow to oversee the growing number of training tasks by restructuring the training workflows as directed acyclic graphs
Participated in creating integration test suites using JMeter for training workflows and key features, ensuring smooth execution prior to build releases.
Was a key enabler for collaboration among cross-functional teams that include application engineers, system engineers, and QA engineers.
Software Engineer
KLA-Tencor
07.2017 - 03.2021
Refactored a legacy C++ monolith into a .NET core microservice for exporting klarf (.klarf is a standardized format of results from a defect inspection tool used in the semiconductor industry)
Created an installer utilizing C# and InstallShield that features centralized JSON configuration management for distributed services, which effectively facilitated the activation of new builds and cut the installation time in the field by an hour.
Demonstrated initiatives in troubleshooting, with a talent for addressing potential issues concerning Linux/windows systems, networking, and Python/GPU environments.
Internship
KLA-Tencor
11.2016 - 05.2017
Developed a scalable monitoring solution for distributed services to oversee all software and hardware alerts (incorporating alerts from OpenStack)
The monitoring system operated as a Windows service while utilizing RabbitMQ for the message queue. High priority alerts were shown as pop ups that required human intervention, leading to a notable decrease in crashes and an increase in reliability.
Internship
KLA-Tencor
05.2015 - 11.2015
Developed a novel image classification engine that differentiates nuisances from defect utilizing images from a scanning electron microscope (SEM).
DataLoader(WCF, .Net) and ImageViewer(WPF) were optimized to load the 500K images and display 1K images per 2 seconds improving the user experience radically.