Mark Sui

Experience

Work Experience

• Research Fellow – Handshake MOVE AI Program
  Jun 2025 – Present · Remote
  Selected to participate in the MOVE AI program, contributing to safety and reliability research for large language models. Designed adversarial prompts and jailbreak strategies to stress-test OpenAI’s Agent model in biosafety and information-hazard scenarios. Conducted systematic QA evaluation of LLM outputs in Project Castor, focusing on contextual accuracy, factual grounding, and mitigation of harmful completions. Collaborated with a cross-disciplinary team to produce reports on AI system vulnerabilities and recommendations for risk reduction. Technologies: Python, Prompt Engineering, Large Language Models, Evaluation Frameworks
• Embedded Systems Intern – Shandong Haike Information Technology
  Jun 2024 – Dec 2024 · Hybrid
  Participated in the development of a radar-driven IoT system for elderly health monitoring. Worked with engineers to integrate mmWave radar modules that could track human movement patterns, restroom usage, and respiration signals without relying on cameras. Conducted sensor calibration and tested system accuracy under different household conditions, including low-light and high-noise environments. Delivered weekly reports on system stability and explored trade-offs between resolution, cost, and privacy protection. Technologies: ESP32, mmWave radar, Python data analysis
• Software Development Intern – Liqun Group
  Jun 2023 – Sep 2023 · Qingdao, China
  Contributed to the design of a WeChat-based social networking app that combined messaging, secure payments, and local community features. Helped redesign the front-end interface for smoother user interaction, and implemented secure payment API integration. Assisted in optimizing the friend-matching algorithm by tuning similarity weights on profile attributes, which improved match accuracy by ~20%. Conducted performance testing that reduced app response latency by ~15%. Technologies: WeChat Mini Program, JavaScript, MySQL
• Student Staff – UC Davis Housing & Dining Services
  Jun 2022 – Nov 2022 · Davis, CA
  Supported dining operations during high-demand periods, coordinating with a team of 20+ staff. Assisted in inventory management, food service, and compliance with health regulations. Gained practical experience in teamwork, problem-solving under pressure, and service operations in a fast-paced environment.
• Tutor – Kumon Learning Center
  2019 – 2021 · Yorba Linda, CA
  Tutored 50+ students across elementary to high school levels. Designed individualized study plans and explained advanced math concepts in accessible terms. Helped raise student performance on standardized assessments by ~25% on average. Mentored younger students on developing consistent study habits and problem-solving strategies.

Academic Projects

• Indoor Environmental Monitoring System – UC San Diego
  Spring 2025 · La Jolla, CA
  Designed and prototyped an ESP32-based environmental monitoring system capable of measuring CO₂, PM2.5, VOCs, temperature, humidity, and noise. Integrated multiple sensors (BME680, SCD4x, PMSA003I, ICS-43434) on a custom PCB, with data displayed on a TFT screen. Implemented a multi-page navigation interface, power management, and data logging features for long-term deployment. Technologies: ESP32, Arduino, I²C/I²S/SPI, PCB Design
• Deep Learning for Climate Emulation – UC San Diego
  Spring 2025 · La Jolla, CA
  Participated in a Kaggle-hosted competition on climate model emulation. Built CNN, Vision Transformer, and MLP architectures to emulate global temperature and precipitation fields using CMIP6 SSP datasets. Benchmarked different models and achieved Top 10 finish (#9/83) with a fast and accurate MLP approach. Technologies: Python, PyTorch, Vision Transformer, Kaggle
• IoT Smart Agriculture Project – National University of Singapore
  Fall 2024 · Singapore
  Developed an IoT prototype for fruit quality assessment using ESP32, camera module, and microphone. Collected tapping sound and visual data, then applied CNN + LSTM models for sweetness prediction. Reached high prediction accuracy while demonstrating non-destructive food quality assessment methods. Technologies: ESP32-CAM, Python, ResNet, LSTM
• FPGA-Based Viterbi Decoder
  Advanced Digital Design · UC San Diego
  Designed and implemented a convolutional encoder and Viterbi decoder on FPGA using SystemVerilog. Built a Mealy FSM-based encoder with recursive, systematic encoding at a rate of 1/2. Developed a decoder architecture with Branch Metric Calculation (BMC), Accumulated Cost Selection (ACS), and Traceback Unit (TBU) for accurate signal reconstruction. Technologies: SystemVerilog, FPGA, Digital Design
• Cryptography Seminar & Research – UC San Diego
  2024 · La Jolla, CA
  Conducted research on lattice-based cryptography in MATH 187B. Delivered a one-hour seminar on Babai’s Algorithm for the Closest Vector Problem. Studied and implemented NTRU signature schemes, analyzing their security against lattice reduction attacks. Technologies: Python, SageMath, Cryptography
• Reinforcement Learning: Blackjack AI
  Spring 2025 · UC San Diego
  Implemented Monte Carlo, Temporal-Difference, and Q-learning methods to train an AI agent for Blackjack. Conducted large-scale simulations (1M+ games) to benchmark convergence of strategies. Achieved ~41% win rate against fixed policies and visualized training process with custom GUI. Technologies: Python, NumPy, RL Algorithms
• Monte Carlo Tree Search Gomoku AI
  Spring 2025 · UC San Diego
  Built a full Monte Carlo Tree Search pipeline (selection, expansion, simulation, backpropagation) for a Gomoku-playing AI. Tuned search iterations (1000–6000) and implemented heuristic pruning, outperforming random baseline significantly. Technologies: Python, Game AI
• Weather Forecasting and Air Quality Prediction
  Data Analysis and Inference · UC San Diego
  Developed machine learning models using Random Forest and Polynomial Regression to predict precipitation and air quality in San Diego County. Conducted feature importance analysis to identify key environmental factors, improving accuracy of AQI predictions. Processed and integrated large meteorological and air quality datasets to provide actionable insights for environmental monitoring. Achieved an R² of 93.2% and MSE of 1.41 for AQI prediction using Random Forest, significantly outperforming Polynomial Regression. Technologies: Python, Scikit-learn, Data Analysis
• Undirected Graph Analysis
  Data Structures and Algorithms · C++
  Built an undirected graph class with support for graph construction, shortest path computation, and connected components detection. Implemented BFS for traversal and connectivity analysis, and applied Dijkstra’s algorithm for efficient shortest path finding. Technologies: C++, Graph Algorithms, BFS, Dijkstra
• UCSD Global TIES Program
  2023 – 2024 · La Jolla, CA
  Joined a multidisciplinary engineering team to co-develop solutions for non-profit partners. Assisted with prototyping, testing, and documentation of technical components, applying embedded systems expertise to real-world challenges. Focus: Community-based engineering, teamwork, embedded systems
• Image Classifier
  Engineering Problem Solving · MATLAB
  Built a versatile image classification tool using AlexNet, with user-selectable options and enhanced accuracy through pre-processing and augmentation. Added remote monitoring capabilities via ThingSpeak integration and Dropbox API for cloud-based storage. Technologies: MATLAB, AlexNet, Image Processing
• Battleship Game
  Software Development & Object-Oriented Programming · C++
  Developed a Battleship game in C++ with an AI opponent, demonstrating inheritance and polymorphism. Integrated a Linux terminal-based user interface and reengineered the game architecture to showcase advanced OOP concepts. Technologies: C++, Object-Oriented Programming, Linux
• Doubly Linked List
  Software Development & Object-Oriented Programming · C++
  Designed and implemented a templated doubly linked list class from scratch, supporting both forward and reverse iterators. Enhanced functionality with additional members and methods to improve efficiency and versatility for data management. Technologies: C++, Data Structures, Algorithms
• AVL Tree
  Data Structures, Algorithms & Programming · C++
  Designed and implemented a self-balancing AVL tree from scratch to support efficient insertion, deletion, and retrieval operations. Optimized balancing mechanisms to reduce search time complexity and ensure consistent performance. Technologies: C++, Data Structures, Algorithms
• Team Leader – Yorba Linda High School Robotics Club
  2018 – 2020 · Yorba Linda, CA
  Led the robotics team in designing and programming robots for competitions. Utilized MIT App Inventor for mobile app integration and organized school-wide robotics events.