Ruey-Tzer (Reggie) Hsu

ad1r1m@r.postjobfree.com github.com/reggiehsu111 linkedin.com/in/reggiehsu +1-669-***-**** Ann Arbor MI, 48105 EDUCATION

• University of Michigan (UMich) - Masters in Computer Science and Engineering 2022.08 - ongoing

• National Taiwan University (NTU) - Bachelor of Science in Electrical Engineering / Double major in Finance 2016-09 - 2021-06

Highlight Courses

Database Management Systems Advanced Compilers Operating Systems Distributed Systems Computer Vision Networking and Multimedia Lab Internet Integrated Circuits Design Computer Architecture Web Programming Machine Learning Applied Parallel Programming with GPUs (Aug. 2023) Advanced Database Systems (Aug. 2023) HCI (Aug. 2023) Leadership and Teamwork

NTUEE Students Association Director of Marketing NTU Artfest Curator EMPLOYMENT Software Development Intern Statementdog 2023-06 - 2023-08 I used Langchain, LlamaIndex, ChromaDB, OpenAI APIs

• Build an indexing system and knowledge graphs from earnings call transcripts using Langhchain and LlamaIndex with OpenAI APIs. Test efficacy using prompt engineering, chain and agent designs.

• Build a proprietary module with async features to replace Langchain and LlamaIndex framework.

• Research retrieval methods using vector stores and knowledge graphs. Backend Development Intern Appier 2022-03 - 2022-07 I used Python3(Django), Kubernetes, Nginx, AWS Services(S3, SQS), Kafka, MySQL, Docker, ELK Stack, Grafana

• Connected local API servers to remote environments with ssh tunneling to access VPC servers. Reduced API backend development time by around 12.5%.

• Migrated servers in a production environment from EC2 to k8s and configured Nginx to retire old services with zero downtime.

• Redesigned backward-compatible API flows to remove local file dependencies.

• Completed 29% backend-related tickets for customer service squad during Q2, which includes designing data schema and migrating MySQL data in production.

Software Development Intern Cloudmosa 2020-07 - 2021-02 I wrote C++ and React Native(Javascript Runtime) code

• Designed a file drag-and-drop feature on Puffin Browser (Chromium based remote browser) with user behavior aware data flows.

• Relocated remote syncing version of LevelDB on client side to resolve reconnection issues using synchronization methods. I used callback functions instead of semaphores or conditional variables that resulted in sporadic deadlocks.

• Implemented browser site settings UI with react native on Windows. Software Development Intern Microsoft 2019-04 - 2020-06 I wrote Python(Pytorch, Flask) and Javascript(React) code

• Built a website that tracks COVID-19 patients’ geo-locations and contact history for Taiwan CDC using React and Azure services.

• Create deep learning models with Pytorch to meet real-world needs, e.g., improving inventory turnover for convenience stores.

• Trained and deployed real-time object detection systems using Yolo-v3 on projects such as faulty screws detection on MRT railways.

PROJECT HIGHLIGHTS AIQUA Corporate product

A personalization platform using AI to manage marketing campaigns and recommends merchandise to promote to groups of users.

Puffin Remote Browser Corporate product

Cloudmosa product that splits chromium code into server and client side, with end users on platforms including Windows, Linux, Android, iOS, MacOS and chrome browser. Backend written mainly in C++, frontend depends on the platform.

Crosstalk Startup product

Video conference web app with collaborative whiteboard to provide immersive user experience. Using agile development and host services in ngrok, Firebase and Agora. Gradually migrate from Agora to WebRTC in production.

Realtime AI Guitar Effector Startup product

I trained a Wavenet-based model to emulate guitar effects and deploy it on Google Coral to achieve real time response.

Energy-Efficient TVM Compiler Advanced Compiler Course Project Add features to the Apache TVM compiler to optimize energy usage. Distributed Key/Value Datastore Distributed System Course Project Four versions of fault-tolerant distributed K/V datastores written in Go: 1. Primary-Backup architecture with a view service server to keep track of server configurations. 2. A Paxos library for distributed consensus, and allows clients to communicate with arbitrary replicas. 3. Consistent hashing server similar to Dynamo to allow horizontal scaling of K/V datastores. 4. Sharded K/V datastores with Paxos groups.

Operating System Operating System Course Project

Thread Library: A single CPU implementation of threads, mutex, and CVs utilizing interrupt disables written in C++. Kernel Memory Manager: A Kernel Memory Manager that implements shared pages with copy-on-write, written in C++.

FileSystem WebServer: A C++ WebServer that supports efficient Read/Write/Delete/Create files and data blocks. Android Malware Detection with Streamed NN School Project Automate Android sandboxes to run deep learning models to identify Android malware apps by analyzing static description files and Internet packages and temporal user data extracted from sandboxes. ECC encoder/decoder Hardware Project

Verilog project encoding 32-bit elliptic curve on prime field with speed-up features using lookup tables. The actual IC chip is produced and sent back to run verification tests. RESEARCH & PUBLICATIONS Research Assistant NTUEE 2020-09 > 2021-06

• MobiSys 21 Poster AR game traffic characterization: a case of Pokémon Go in a flash crowd event

• AINTEC 21 Best Paper Twitch’s CDN as an Open-Population Ecosystem KEY SKILLS Languages • C/C++ • Python (Pytorch) • Go • Javascript (Node.js, React, Nextjs) • Java • HTML5, CSS • R • Verilog Tools • Git • llvm • Kafka • k8s • docker • MySQL • Unix • Redis • Grafana • ELK Stack • Azure Services • AWS Services (S3, SQS) • Firebase • ngrok • Nginx • CUDA