Mar 2023 – Dec 2023 · 9 mos · San Diego, California, United States

• ◦ CR Automation Platform: Design a system to sync Customer Report (CR) database to edge server for faster access, downstream apps and ML training. Design and develop HTTP API generator accessing remote database and scheduling runner on edge server. Develop a demo web app with React.js + Flask. Deploy and config MySQL database.
• ◦ Wireless communication: Develop features on Modem-AP interface (MIPC) with C for customization. Develop and test with Python host library on test board. Investigate wireless technics and internal implementation from top-down.
• ◦ Accelerate query speed from 40-60s to less than 10s. To scale up with global group accessing with 3k+ engineers.

Jun 2022 – Mar 2023 · 9 mos · Mountain View, California, United States

• ◦ Machine Status Update Service: Created a bot status management gRPC service to update status from Autonomous Vehicle when offloading disk to Google Cloud SQL Storage with PostgresSQL engine. Resolved compatibility problem in previous system depending on individual infrastructure onboard and extended flexibility to choosing between remote update and local fallback. Create unit test and integration test in Jenkins CI/CD pipeline.
• ◦ Google Transfer Appliance Offloading: Designed a network-agnostic disk-offloading pipeline based on Google Transfer Appliance. Integrated the pipeline with existing CI pipeline(BuildKite). Created monitoring and metrics based on Google Cloud Platform.
• ◦ Remove 2 sets of onsite infrastructure, which lower the maintenance cost for onboard software. Boost the log onboarding process by 10* under poor network connection.

Apr 2021 – Aug 2021 · 4 mos · Beijing, China

• ◦ Secret Injector Service: Created a user token management system enabling automatic token injection to Kubernetes container via Hashicorp Vault. Deployed web hook trigger to automatically launch sidecar container for injection. Implemented Kubernetes API agent for related K8S resource management.
• ◦ ABAC Authentication & Authorization Design: Promoted the current identity system from simple authentication to a centralized authorization and permission administration service. Investigated standard authentication protocol including OIDC & SAML. Designed Attribute-Based Access Control policy system for infrastructure applications, restricting access to frontend components, backend API and backend database objects.
• ◦ Identity System Demo Development: Implemented a feature demo constructed by Single-Page Application with Vue and backend service with Django.

Aug 2020 – Mar 2021 · 7 mos · Beijing, China

• Honored "Star of Future"
• ◦ Project – Automatic Model-Launching System: Change the pipeline of implementing new Deep Learning models from purely manual work to automatic pipeline.
• ◦ System Design: Designed an automatic deployment system for Deep Learning framework enabling both real-time call and asynchronous process with input uploaded and stored.
• ◦ Multi-Media Processing: Implement multi-media pre- and post-processing including video segmentation & combination, resizing, frame feature extraction via FFMpeg
• ◦ Content Management & API: Designed & tested RESTful API for resources to communicate with front-end service. Implemented a metadata management system enabling CRUD with MySQL. Deployed OSS for file storage.
• ◦ Inter-Service Communication: Implement gRPC API between micro-services. Enabled robust message communication module for asynchronous tasks via Kafka data streaming.
• ◦ Reducing the process of real-life service implementation of newly promoted multi-media processing Deep Learning frameworks from 1 month to less than 1 week.

Feb 2020 – Jul 2020 · 5 mos · Beijing, China

• ◦ Snippets Design: Setup snippets based on multi-platform applications e.g. iOS, Android, H5 etc. and correspond quality assurance.
• ◦ RDBMS Feature Design for Structural Data: Designed implemented CRUD strategies of multiple structural data in RDBMSs (Hive & MySQL).
• ◦ Data Analysis: Analyzed client behavior of 100k+ users via HiveSQL and designed and implemented auto-generated charts for daily report via Clickhouse.
• ◦ The designed feature has extended the content category system from single layer to multiple with similar response time.

Jun 2019 – Oct 2019 · 4 mos

• ·Implementation & Modifications on various baseline Seq2seq models
• ·Data Analysis on mainstream Neural Machine Translation tasks
• ·Design & implementation of a novel Meta-Learning related translation pipeline based on FAIRSeq