ML Platform - Inference Platform

Software Engineer in Google Gmail Backend

● Rewrote Gmail Indexing Library for new storage system, streamlining the Gmail indexing process across various clients and enhancing efficiency.

● Implemented support for both native indexing and relevancy indexing, resulting in improved resource management and enhanced flexibility for reindexing operations.

● Participated in multiple projects focused on rolling out the new Gmail Relevancy search and indexing. Made substantial contributions to enhancing the efficiency, reliability, and completion percentage of the workflow for reindexing relevancy ranking data.

● Provided support for analyzing and debugging relevancy search issues and ranking regression. Successfully scaled the user base of the new Gmail Relevancy search from a couple of million to over 150 million users.

● Implemented, deployed, and maintained a high-capacity key Server for Gmail Delivery, optimizing the email migration process for large clients. This implementation increased the email batch upload capacity from 1-2 emails/second to a minimum of 10+ emails/second, ensuring a smooth transition to the Gmail system.

● Developed some key features that are responsible for handling Gmail threading, de-duplication spam/phishing signal generation, and support for smart feature computation, contributing to an enhanced user experience.

● Conducted in-depth analysis of Gmail delivery system traffic, resulting in the reduction of unnecessary RPCs, payload, and network bandwidth. This optimization effort saved approximately 7 SWE resources per year.

● Worked on improving the stability and reliability of Gmail Delivery Pipeline. Implement features such as differentiated network priorities for delivering emails based on user activity levels, resulting in improved general delivery latency for end users.

• Beepi Search 2.0, worked on developing new Beepi Search based on Elastic Search and Java. Using netty.io to implement asynchronous web server and using ELK stack to monitor whole System.

• Autosuggestion, designed and implemented autosuggestion system for search, showing ranking results based on popularity score from history. In addition, it supports synonyms, misspellings and showing similar suggestions.

• Fields Ops Scheduler, using OptaPlanner with simulate annealing algorithm to solve field-ops task scheduling problem. Taking Google real-time traffic information in cost formula, the project will optimize the routes scheduling for specialist.

• Paperwork, worked on developing an automated paperwork system to automatically fill out all needed documents during transaction and implemented an asynchronous server with Tornado to handle all printing requests, then using Google Cloud Printing to print remotely. (Python)

• Grader of Graduate Level Course: Fundamental Algorithms (CSCI-GA.1170-001)

• Grade assignments and answer questions