- Led and managed a 5-person team, overseeing driver assignment optimization and ETA estimation.- Implemented driver allocation strategies, such as multi-vehicle, back to back, allocation based on food prep time, zone-based, self-pickup and scheduled delivering, resulting in increased order fulfillment rates.- Developed and implemented a real-time ETA prediction engine from scratch, utilizing a graph-based computing strategy.- Achieved sub-200ms latency for 95% of estimates during peak hours, as measured by application performance monitoring and maintained an availability rate above 99.99%.

- Led a 5-person team in managing the company's user center, authorization, and SSO systems.- Achieved a significant increase in service availability, boosting uptime from 99% to 99.99% through strategic measures to reduce downtime.- Successfully optimized SQL, memory, and middle-ware processes, resulting in a remarkable decrease in average response time to 50 ms.- Developed and implemented a highly efficient near real-time personnel data synchronization system, processing updates from HRIS within an impressive 100 milliseconds for 99% of 1500 daily changes.- Created an open platform that facilitated dynamic configuration of multiple authoritative data sources, enhancing flexibility and adaptability.- Spearheaded the design and development of a cutting-edge RBAC authorization system in Java, replacing the previous static configuration and enabling automatic permission changes in under 1 second for 95% of 10,000 daily user updates.- Implemented a pattern-based authorization platform, granting roles to users across 40 internal systems, ensuring seamless access control.

- Designed and implemented an Apache Mesos-based elastic container cloud platform, achieving a deployment time of less than 10 seconds. The platform successfully supported over 1800 concurrent Docker containers across 50 compute nodes, with a node self-discovery and self-healing uptime of 99.99%.- Developed a non-invasive log collecting and monitoring system for containers, enabling custom filtering and actions on predefined log patterns.- Built a container performance monitoring and alerting system that fetched metrics from running containers every second, with an average retrieval time of 50 milliseconds. Customized Grafana dashboards and implemented a rule engine based on Bosun to detect and notify on real-time resource usage outliers.- Designed and developed a real-time advertising delivery system using the Storm framework. This system successfully pushed targeted ads to user TVs within 500 milliseconds for 95% of events, as measured by elapsed time for 10,000 ad requests. The Java/Storm system processed over 5,000 ad requests per second based on users' online behavior events against configurable rules. This system got prioritized for the overall video monetization strategy of the company.

- Provided expertise in X86 hardware enablement and troubleshooting, with a specific focus on BIOS/ACPIfor Linux kernel to enable hotkeys, brightness adjustment, and resolve power management issues on PC/laptop platform.- Successfully enabled Ubuntu on ARM-based thin client hardware, ensuring smooth functionality and compatibility. Assumed responsibility for maintaining the entire Kernel tree, including reviewing patches,backporting, building, and distributing updates.- Facilitated IPv6 network stack enablement for custom devices, guaranteeing reliable passage of IPv6Ready Phase 2 testing.

I/O subsystem of Internal Linux Kernel* Analyze the requirement for I/O of products, optimize its performances.* Follow vendors and kernel upstream, update and optimize internal I/O drivers.* Do some modifies to I/O drivers and block layout to fit our needs.* Writing some user-space tools for devops such as disk location and work load tracing.Hard Disk Failures Predict and Repair* Research the ways to repair a failure-like hard disk and how to predict its failure.* Train a ML classifier using S.M.A.R.T data of disks to predict a hard disk failure, AUC to 0.95.* Design and implement a system that runs on every internal server to automatically find a pre-failure hard disk and try to fix it, If the fix fails, automatically report to replace it. The total fix rate is up to 0.6.* Design and implement a distributed system collecting S.M.A.R.T information from every hard disks in every servers for modeling training and failure predict.Low-cost, Low-power server* Analyze the internal usage and requirement for Low-cost, Low-power server.* Together with vendors, design specifications for ARM and Tilera servers.* Do performance tests for ARM and Tilera servers, maintain these 2 development servers.* Port internal programs to ARM architecture.

Coordinate with other teams to setting up test environment. Write test plan, design new test cases and write code using C, Bash and Python. Automate existed manual test cases. Responsible for maintaining automatic scripts, analyze result and track bugs.Fix some bugs of nfs-utils and linux kernel.

Get familiar with the whole coreboot project, including the source code and the building system. Finish porting coreboot to ARM architecture, designed the ROM structure of coreboot on ARM from scratch. Take charge of the whole source code version are controlled under git, and programmed by C and ASM language. Effective communication with USA and German developers by email and IRC.