Email: | GitHub: yangtaige

Education

University of California, Berkeley — Visiting Student, Computer Science Aug 2025 – Present
GPA: 4.0 / 4.0

ShanghaiTech University — B.Eng. Biomedical Engineering (in progress) Sept 2023 – Present
GPA: 3.81 / 4.0
Minor in Information Technology
Honor: Outstanding Student, 2023–2024

Publications & Manuscripts

  1. Yiwei Hou, Hao Wang, Muxi Lyu, Marius Momeu, Eric Nguyen, Taige Yang, Dawn Song, Koushik Sen, David Wagner. “Revelio: Cost-Efficient Agentic Memory Safety Vulnerability Detection Across Repository-Scale Codebases”. Under review at IEEE Symposium on Security and Privacy (S&P), 2026.

Research Experience

Computational Imaging Lab, UC Berkeley — Undergraduate Researcher Feb 2026 – Present
4D Digital Inline Holographic Particle Velocimetry
Advised by Dr. Nalini Singh & Mingxuan Cai

  • Adopted a Lagrangian representation with floating-point coordinates for off-grid particle trajectory tracking.
  • Developed an analytical forward projection model and Angular Spectrum Method (ASM) initialization for holographic reconstruction.
  • Represented 4D velocity fields via physics-informed MLPs to regularize ill-posed inverse problems in fluid dynamics.
  • Built an end-to-end pipeline integrating particle localization and 4D flow field reconstruction.

Sky Computing Lab, UC Berkeley — Research Assistant Mar 2026 – Present
Zero-Day Vulnerability Detection via Revelio
Advised by Yiwei Hou

  • Benchmarked the Revelio framework’s efficacy in detecting memory vulnerabilities in V8 and libtorrent.
  • Engineered Dockerized fuzzing environments with OSS-Fuzz and LLVM sanitizers (ASan, MSan, UBSan).
  • Analyzed LLM-generated Proof of Concept exploits to assess the precision of agent-based security auditing.

Translational Neuroimaging Lab, ShanghaiTech — Undergraduate Researcher Aug 2024 – Nov 2025
Sleep Stage Classification via Mouse fMRI
Advised by Prof. Zhiwei Ma & Yiyun Qi

  • Applied spectral clustering to parcellate brain regions and extracted dynamic Functional Connectivity (dFC) features using sliding-window correlation matrices from fMRI signals.
  • Trained SVM, Random Forest, and MLP models to classify NREM/REM sleep cycles using ECoG as ground truth.
  • Identified subcortical biomarkers tightly coupled with sleep state transitions via cross-modal correlation analysis.

Project Experience

CS162: Operating Systems — UC Berkeley Spring 2026
C · Rust

  • Co-developed a monolithic Pintos OS kernel: implemented file operation syscalls, ROX memory protections, user-kernel boundary security check, 1:1 pthread-to-kernel-thread mapping, pthread_join/pthread_exit syscalls, kernel-backed mutexes and semaphores, exit/wait process control syscalls for multi-threaded processes, and a 64-entry buffer cache with N-th Chance clock eviction.
  • Built a fault-tolerant distributed MapReduce framework with heartbeat-based failure recovery, and engineered a highly available distributed Key-Value store based on the Raft consensus algorithm.
  • Engineered a concurrent HTTP/1.0 server using a custom thread pool, a POSIX-compatible shell with full job control and pipelines, and a malloc/free allocator with explicit free-list and block coalescing.

CS152: Computer Architecture and Engineering — UC Berkeley Spring 2026
C · RISC-V Assembly

  • Reverse-engineered cache hierarchy (L1: 16 kB, 64 B line, 8-way) via microbenchmarks and hardware performance counters.
  • Co-architected an Alpha 21264-style tournament branch predictor (4096-entry global + 1024-entry local + arbiter).
  • Vectorized complex arithmetic and DGEMV in RISC-V Vector Assembly.
  • Optimized multithreaded matmul achieving ~7.0× speedup (MI) and ~2.2× speedup (MSI) via matrix pre-transposition and cache-coherence-aware memory access.

CS161: Computer Security — UC Berkeley Fall 2025
Go · C · x86 Assembly

  • Architected a cryptographically secure distributed file-sharing system in Go using AES Encrypt-then-MAC, RSA hybrid encryption, and HashKDF-derived keys; implemented strict key-regenerating revocation.
  • Exploited memory vulnerabilities (buffer overflows via sign-extension bugs, SFP pivots, stack canary leaks, ASLR bypass via printf info-leak, format string %hn writes, ret sleds, TOCTOU race conditions).
  • Discovered and chained SQLi, stored/reflected XSS, CSRF, and directory traversal vulnerabilities in a mock web platform.

CS61C: Great Ideas of Computer Architecture — UC Berkeley Fall 2025
C · RISC-V Assembly

  • Implemented a Snake game engine in C with dynamic memory management verified by Valgrind Memcheck.
  • Hand-wrote a neural network forward-pass classifier in pure RV32I assembly with strict calling convention.
  • Designed a two-stage pipelined RISC-V (RV32I) processor in Logisim with forwarding and hazard stall logic.

CS188: Introduction to Artificial Intelligence — UC Berkeley Fall 2025
Python · PyTorch

  • Implemented A* search with custom admissible heuristics, Alpha-Beta Pruning Minimax, and Expectimax agents.
  • Built Q-Learning and Approximate Q-Learning agents with ε-greedy exploration; solved MDPs via Value Iteration.
  • Engineered HMM-based ghost tracking with particle filtering; built a character-level GPT with attention mechanisms.

BME1312: AI in Medical Imaging — ShanghaiTech Spring 2025
Python · PyTorch

  • Reconstructed dynamic cardiac cine MRI from 5× undersampled k-space data using a 3-stage cascaded 3D U-Net with CBAM attention and Data Consistency layers; achieved PSNR 37.07 dB and SSIM 0.9626.
  • Developed 2D cardiac segmentation models (U-Net, MedNeXt, MMNet); achieved Dice 0.9611 on the left ventricle.
  • Analyzed CE vs. Dice loss gradient formulations and applied combined dynamic-weighted loss for stable convergence.

BME1107: Biomedical Electronics II — ShanghaiTech Spring 2025
Cadence

  • Designed a complete EEG signal processing module: folded-cascode OTA preamplifier (<30 µW, maps input to 0.7–1.6 V), sampling circuit, single-slope ADC with high-linearity ramp generator, and DAC.
  • Suppressed end-to-end latency to <1 µs by aligning the ramp reference with preamplifier output and increasing CLK to 500 kHz.

BME1109: Biomedical Electronics I — ShanghaiTech Spring 2025
PCB · Python

  • Designed and assembled a fully functional PCB-based ECG acquisition circuit, tested on human subjects.
  • Developed a Python program for real-time electrocardiogram visualization and heart rate display.
  • Collaborated with 2 classmates.

Laboratory Experience

BME1108: Human Anatomy and Physiology — ShanghaiTech Spring 2025

  • Conducted hands-on cadaveric dissection observation, examining skeletal structures, musculature, internal organs, and the peripheral nervous system.

PHYS1113: General Physics II Lab — ShanghaiTech Fall 2024

  • Performed experiments covering electromagnetism and optics.

CHEM1103: General Chemistry I — ShanghaiTech Fall 2024

  • Performed experiments in structural chemistry.

PHYS1111: General Physics I Lab — ShanghaiTech Spring 2024

  • Performed experiments covering kinematics, dynamics, and thermodynamics.

Skills

Programming: Python · C · Go · Rust · RISC-V Assembly · x86 Assembly · MATLAB · SQL · C#

Hardware: Cadence (analog IC design & simulation) · PCB design and debugging

Tools: Git · PyTorch · GDB · Valgrind · Docker · Logisim · Unity

Domain Knowledge: Computer Security · Operating Systems · Computer Architecture · Machine Learning & AI · Medical Imaging · Computer Networks · Signals and Systems · Biomedical Engineering · Human Anatomy and Physiology