SOFTWARE ENGINEER
Tianrui Xia
ResearcherProblem SolverBuilder
MS student in Computer Science at University of Pennsylvania. Specializing in distributed systems, cloud infrastructure, and deep learning. Previously at Apple, Microsoft Research Asia, and Scaleflux.
About Me
Education
University of Pennsylvania
MS in Computer Science
GPA: 4.0
University of Southern California
BS in Computer Science
GPA: 3.98
Skills & Expertise
PROGRAMMING LANGUAGES
C/C++PythonJavaJavaScriptGoSQLMATLABHTML
TECHNOLOGIES & FRAMEWORKS
PyTorchPrometheusGrafanaCI/CD
AREAS OF EXPERTISE
Distributed SystemsCloud InfrastructureDeep LearningMachine LearningNetworked Systems
Experience
Software Engineer
Apple Services Engineering
- ▹Improved the scalability and reliability of Apple's cloud compute service, a platform supporting core initiatives such as hardware and AI/ML developments
- ▹Designed the automatic agent for the cluster scalability testing tool, reducing manual effort in cluster setup, initiation, and workload submission
- ▹Created dashboards for centralized performance comparison by summarizing and publishing performance metrics with Prometheus.io and Grafana, enabling faster detection of performance regressions and scaling bottlenecks
- ▹Built a CI/CD pipeline to run large-scale simulations daily, ensuring continuous validation of cluster performance
Firmware Engineer
Scaleflux
- ▹Performed Quality of Service testing on the latest firmware using flexible I/O test tool; analyzed logs to identify root causes of failures and performance drops
- ▹Tuned parameters to enhance I/O performance across varying read/write patterns on ARM architecture
- ▹Boosted sequential I/O consistency from 60% to 90% by mitigating speed drops caused by erase operations blocking reads. Adjusted preemptive Block-to-NAND allocation for garbage collection, resolving instability in random read/write tests and increasing overall performance from 70% to over 95%
Undergraduate Researcher
Networked Systems Lab - Volumetric Video (USC)
- ▹Awarded the CURVE fellowship by USC Viterbi School of Engineering; Participated in the volumetric video project
- ▹Searched and tested different models and codes for foreground-background segmentation to increase the performance and reliability of volumetric video streaming
- ▹Explored the possibility of utilizing NeRF (Neural Radiance Fields) to stream 3D scenes in real-time; Tested the performance in terms of training time and rendering quality with pictures focusing on human
- ▹Applied segmentation masks to the NeRF models and observed the performance, including the training time and quality
Research Internship
Microsoft Research Asia (MSRA)
- ▹Coauthored the paper: "Empowering In-Browser Deep Learning Inference on Edge Through Just-In-Time Kernel Optimization". Accepted for presentation at the 22nd ACM International Conference on MobiSys 2024
- ▹Conducted comprehensive experiments on the 3070-ti GPU, assessing Matrix Multiplication speeds under diverse configurations and delving into performance variations
- ▹Proved the superiority of our kernel searching policy against notable benchmarks like tf-js, autotvm, and ort-web
Featured Projects
iFlow
Max-flow/Min-cut Algorithm Visualizer
Published at SIGCSE Technical Symposium 2025
- ●Coauthored paper "iFlow: An Interactive Max-Flow/Min-Cut Algorithm Visualizer", accepted by the SIGCSE Technical Symposium 2025 program
- ●Developed an online interactive visualizer for Max-flow/Min-cut algorithms as a group project in the Capstone course
- ●Designed and implemented educational functions under the guidance of Prof. Kempe, including step-by-step self-testing and Min-cut identification, distinguishing our tool from existing visualizers
- ●Distributed user experience survey for Computer Science students at USC, which received positive feedback
Algorithm VisualizationEducational ToolWeb DevelopmentGraph Algorithms
View ProjectPennCloud
Distributed Cloud Platform
- ●Built a distributed cloud platform featuring webmail and file storage services with frontend and backend servers
- ●Implemented a distributed backend key-value store, enabling data replication, fault tolerance, and efficient recovery
- ●Designed and developed frontend server functionalities, including user authentication, webmail, and file storage services, with a multithreaded architecture to handle concurrent HTTP requests
Distributed SystemsKey-Value StoreMultithreadingHTTP
View DemoMaker Projects
Early projects from high school - building, creating, and learning
Drawing Machine
Paper Clock
Rubik's Cube Solver
Get in Touch
I'm always open to new opportunities and interesting projects. Feel free to reach out!
👋
Let's build something amazing together!