Projects

Contributed to the development and scaling of BoxiStore, a product that reached over 300,000 daily active users. Developed RESTful APIs for warehouse automation in Java Spring Boot, improving order processing speed by 40% and reducing errors by 25%. Led the migration from a monolithic architecture to microservices on AWS’s managed Kubernetes, enhancing scalability and reducing operational costs by 35%. Optimized data handling and storage, reducing API response times by 30% and developed observability toolchains for improved monitoring and incident response.

Designed and implemented a hybrid file system combining a local SSD and cloud storage services akin to AWS S3. Featured block-level deduplication based on Rabin Fingerprinting, reducing cloud storage costs by approximately 60%. Implemented a snapshot mechanism for data integrity and failure recovery, and utilized spare SSD capacity for caching, enhancing performance and reducing costs.

Developed TinyKV, a distributed key-value storage system in Golang that integrates the Raft consensus algorithm for fault tolerance and system reliability. Focused on features like membership changes and leadership changes, implemented multi-version concurrency control for transaction management, and developed TinyScheduler for centralized node management and efficient communication using Protocol Buffers over gRPC.

Implemented core components of BusTub, a relational DBMS in C++, including a buffer pool manager for efficient page management and a concurrent disk-backed B+ tree for fast data retrieval. Optimized query processing speed significantly under various workloads by employing fine-grain locking and an effective replacement policy.

Engaged in the development of PINTOS, an operating system kernel, as part of a project at UC Berkeley. Implemented advanced OS features like dynamic memory management, synchronization, preemptive scheduling, and an inode-based file system, achieving top performance in the class.

Developed Netier, an event-driven TCP server framework in modern C++, employing the reactor pattern for efficient event handling and resource management. The framework simplifies the development of high-performance server applications by encapsulating low-level networking operations.

Engineered a secure file sharing system, employing standard cryptographic primitives to ensure file integrity and confidentiality over unsecured channels. The project was a key part of a computer security course, focusing on the application of cryptography in real-world scenarios.

Executed a parallel N-body simulation using various parallel computing methods, including MPI, Pthread, OpenMP, and CUDA. This project, part of a distributed and parallel computing course, showcased the effectiveness of different parallelism techniques in enhancing computational performance on school clusters.