The Legend of Zelda
Full recreation of Zelda's first dungeon in Unity (C#), including all scenes, enemies, and mechanics faithful to the original.
Hello, I'm
Zhongqian (Henry) Duan
- ☁️ Software Engineer at AWS
- 🕹 Ex Software Engineer at Rec Room
- 💻 Ex Machine Learning Engineer Intern at Apple & NIO
- 🔭 M.S. in Computer Science at University of Michigan
- 📚 Dual degree — University of Michigan (B.S. CS) & Shanghai Jiao Tong University (B.S. ECE)
- 🍁 Fun fact: I love 📸 🏓 🍜 🎮
I build reliable, scalable software — from distributed cloud systems to real-time game engines. Currently exploring opportunities in software engineering where I can work on high-impact, user-facing products.
Work Experience
Software Engineer — Amazon Web Services
- Redesigned the Free Zone Allocation strategy in Kotlin to optimize S3 Glacier retrievals, stabilizing read throughput and mitigating tail latency via a fair resource distribution model that enhanced cross-AZ system reliability.
- Architected a standalone storage validation stack on EC2 to verify SCSI I/O operations, eliminating dependencies on legacy network infrastructure and enabling S3 Glacier expansion into new global regions with 100% data reliability.
- Engineered decoupled mapping logic within VirtualTapeHandle to abstract shards from partitions, facilitating adoption of high-density media and next-gen hardware with zero downtime.
- Architected a serverless, event-driven diagnostics framework using Python, Lambda, CloudWatch, and DynamoDB to ingest real-time failure signals — reducing on-call MTTR by 30%.
Software Engineer — Rec Room
- Engineered runtime execution logic for 20+ core components in the CircuitV2 visual scripting engine using C#, extending Turing-complete capabilities available to millions of VR creators.
- Architected a deterministic state-management system for complex circuit graph connections (Undo/Redo), ensuring atomic restoration of user actions and establishing an extensible API pattern for future tools.
- Developed a networked visual effect system (Motion Trails) using Protobuf for serialization and Photon for real-time sync, optimizing bandwidth in high-concurrency multiplayer environments.
Machine Learning Engineer Intern — Apple
- Built distributed training infrastructure using Python and gRPC, enabling multi-GPU training and centralized data serving — increasing data throughput by 20%.
- Developed a containerized simulation data pipeline with Docker and S3, supporting real-time reinforcement learning workflows.
- Trained a foundation model for FEA simulations in PyTorch, achieving a 5% reduction in Mises Stress Loss.
Machine Learning Engineer Intern — NIO
- Optimized production ML pipelines for 3D object tracking using PyTorch, achieving a +4% increase in F1 score by refactoring deep learning architectures for better feature extraction.
- Engineered an automated evaluation framework to process 200+ hours of autonomous driving footage, utilizing Mask R-CNN to accelerate dataset validation and reduce manual labeling overhead.
- Proposed a lightweight CNN in PyTorch to remove lens distortion effects.
Graduate Student Instructor — University of Michigan
- Led weekly discussions and office hours for 60+ graduate students in EECS 587 Parallel Computing, covering CUDA, OpenMP, and MPI.
Selected Projects
Game Development
Molecule Dash
Original 2D survival game in Unity (C#) — navigate a molecule through darkness while dodging waves of virus attacks.
Asylum 7
3D first-person horror puzzle game in Unity (C#) — survive in darkness using a laser to control lighting, fight zombies, and find your way out. Led game design and wrote the majority of the codebase.
FantasyAR
Full-stack AR fighting game built in Unity with a team of 4 — integrated InteractionSDK for hand tracking, Recognissimo NLP model for voice control, GoMap for real-time AR location minimap, and a Node.js + MySQL backend.
Research
Text-Image Pair Generation via Pre-trained Vision-Language Models
Built a vision-language dataset generation pipeline using ChatGPT and Midjourney, combining Text-to-Image and Image-to-Text tasks to benchmark captioning models — published on EECS 595 Student Project Gallery.
High Resolution Demoire
Designed HRDeNet, a deep neural network pipeline (Downsample → Demoire → Super Resolution) to remove moiré patterns from high-resolution images.
Education
University of Michigan, Ann Arbor
M.S.E. in Computer Science and Engineering | GPA:4.00/4.00 | Aug. 2022 – Dec. 2023
Selected Courses
- Web Systems
- Computer Networks (A)
- Randomness & Computation (A)
- Parallel Computing (A)
- Natural Language Processing (A)
University of Michigan, Ann Arbor
B.S.E. in Computer Science | GPA: 3.9/4.0 | Aug. 2020 – May 2022
Artificial Intelligence
- Computer Vision (A)
- Intro to Machine Learning (A)
- Intro to Artificial Intelligence (A)
- Deep Learning for Computer Vision (A)
Systems & Programming
- Database Management Systems (A+)
- Intro to Computer Organization (A+)
- Foundations of Computer Science (A-)
- Operating Systems (A-)
- Intro to Computer Security (A)
- Applied Parallel Programming with GPUs (A)
Game Development
- Computer Game Design (A)
Shanghai Jiao Tong University
B.S.E. in Electrical and Computer Engineering | GPA: 3.71/4.00 | Sept. 2018 – Aug. 2022
ECE
- Intro to Computer and Programming (A)
- Programming and Data Structure (A-)
- Electromagnetics (A+)
- Circuits (A-)
Mathematics
- Honors Mathematics II (A+)
- Honors Mathematics III–IV (A)
- Probabilistic Methods in Engineering (A)
No. 2 High School of East China Normal University
High School | Sept. 2015 – June 2018
Honors
- 🏅 James B. Angell Scholar — achieved all-A record for 2+ consecutive terms, University of Michigan
- 🎓 University Honors / Dean's List — 4 Terms, University of Michigan
- 🏆 Excellent Scholarship — Shanghai Jiao Tong University
Interest
Photography
Get In Touch
Feel free to drop me a message.