Cloud - Technical Infrastructure - Unified Fulfillment Optimization

SWE data center planning: Google’s data center planning impacts all Google’s internal and external services, providing storage and computing power to support production data processing for flagship products such as Ads, Search, Youtube, and GCP.

Data center rack packing automation: Collaborated with cross-functional teams involving the collection of requirements, design, execution, and testing of a system that significantly streamlined data center rack packing and modeling processes, leading to a savings of thousands of SWE work hours.

Change Management: Executed the design and development of Change Management System within the data center planning pipeline workflow, Effectively managing over $100M worth of hardware. This highly available (99.99%) distributed system reduced the on-call workload and provided guarantees of data integrity.

Team contributions: Collaborated on multiple on-call rotations to ensure seamless service updates and releases, exceeding 99.9% uptime for main services.

• Validated AMD GPU machine learning library (ROCm) to support PyTorch

• Automated PyTorch tests by integrating GitHub repositories to the Jenkins projects with the help of ECR and Docker

• Improved CI build time by parallelizing tests and implementing recovery mechanism using Python

• Reproduced failed test cases and find the regression points with Bash, Git, and Docker

• Wrote scripts to package Python library for ROCm PyTorch

• Contributed to open-source PyTorch repository to solve AMD-related issues

(Docker, Linux, Jenkins, Bash, Jira, ECR, S3)

Face Recognition Sign-in System (Full-stack web application; AWS EC2, S3, AWS Rekognition):

• Integrated a third-party face recognition service to an existing video conference web application using Python, PostgreSQL, and Docker

• Developed Rest API for CRUD operations in Python Flask with PostgreSQL

• Designed data model and query for a relational database

• Deployed applications to a cloud server, AWS EC2, with the help of tools such as Docker and Nginx

• Implemented UI functionalities such as sign-in/sign-up page and pagination using React

• Implemented face recognition as one of the two-factor authentication options

Object Detection & Computer Vision-Based Application (Google Cloud Platform, Compute Engine, Tensorflow, Python, Flask):

• Collected, annotated images, and trained model

• Deployed an object detection machine model to Compute Engine in Google Cloud Platform

• Developed server-side web API with reverse proxy enabled to handle client requests

• Developed front-end application using React.js

Robot Monitoring Application (Python, NodeJS, MongoDB, Linux, ROS, OpenCV, Android - Java):

• Built a full-stack web application to monitor robot status

• Developed gRPC client and server to establish communications between server and robots

• Developed an Android app to enable end-to-end use cases

• Gather sensor data (IMU) and wheel odometry data for sensor fusion

• Improve robot localization using extended Kalman Filter

• Implement object detection (TensorFlow) and distance estimation for a detected object with Intel RealSense D435i depth camera and machine learning techniques

• Analyze and apply different SLAM packages such as Google Cartographer, RTAB-Map, Gmapping

Participated in development of new or existing software to support evolving needs of the project, such as road detection, autonomous car, robotic arm etc.

Details:

• Designed programs that interface with new technologies

• Hands-on experience with testing and debugging the products

• Optimized the performance of a lane detection program by 10% on the AGV-4000 autonomous driving robot

• Developed ABB IRB260 robotic arm application for automation with TCP socket programming