What I care about
I like technical work that combines strong reasoning with real-world usefulness. My interests sit around machine learning systems, HCI accessibility, data-intensive analysis, and security-oriented infrastructure.
Computer Science · UW-Madison
Hi, I’m Siqi.
I build with machine learning, research, cybersecurity, and practical software systems.
I’m a computer science student at the University of Wisconsin-Madison with a strong interest in technically rigorous work that still feels clean, useful, and well designed.
About
Current direction
Right now I'm balancing research, engineering, and operational support. That means building models, working on accessible interaction ideas, and staying grounded in real technical systems and user needs.
What's New
Recent Highlights
Summer 2026
Accepted into Summer of AI Laboratory (SAIL)
Selected for the Summer of AI Laboratory (SAIL) program at UW-Madison, sponsored by the College of Computing & Artificial Intelligence and supported by OpenAI. Over the course of this program, I will be collaborating with peers and industry mentors to design and build innovative, real-world AI applications.
Summer 2026
2026 Central State Section The Combustion Institute Undergrad Research Competition Winner & Presenter
Attended the 2026 Central State Section The Combustion Institute (CSSI) at Purdue University, exploring advanced topics in combustion field.
Spring 2025
URS Symposium Presentation
Presented "Machine Learning-Based Virtual Sensor for Automotive Engine Combustion Monitoring" poster at the UW-Madison Undergraduate Research Symposium and received Certificate of Recognition for outstanding design and scholarly communication (April 2026).
Summer 2026 - current
AR/VR Accessibility Research
Working with mentor on AR/VR systems for low vision accessibility using Meta Quest 3 and Intel RealSense D455.
Experience
Work ranging from engineering and research to operations and support.
Summer 2026 - Present
Engineering
Software Development Engineer
Lead the design and development of end-to-end features for the REC application, a mobile platform connecting users through shared interests, optimizing core user journeys from onboarding to recommendation delivery.
Conducted comprehensive code quality audits and systematic repository refactoring, which resolved critical UI/UX bottlenecks and improved app load times. Managed the complete software lifecycle, including regression testing, continuous integration, and App Store deployment pipelines, while mentoring junior engineers and creating onboarding documentation.
Responsibilities & Impact
- Developed responsive frontend screens and backend API services to enhance onboarding flows and search functionality.
- Formulated and executed testing strategies, reducing user-reported bugs by refactoring legacy modules.
- Directed App Store release cycles, ensuring compliance with Apple's developer guidelines and seamless user transitions.
Summer 2026 - Present
Research
Undergraduate Research Assistant
Design and evaluate intelligent interaction systems for low-vision users in augmented and virtual reality (AR/VR) environments.
Supported accessibility-focused Human-Computer Interaction (HCI) research by conducting literature reviews, structuring user studies, and building prototypes. Developed accessibility interfaces on Meta Quest 3 and Intel RealSense hardware, investigating how spatial audio and computer vision cues can assist visually impaired users in spatial navigation.
Responsibilities & Impact
- Built spatial mapping and feedback prototypes using Meta Quest 3 and Intel RealSense D455 depth cameras.
- Co-authored user study protocols and assisted in executing user evaluation sessions with low-vision participants.
- Analyzed qualitative and quantitative interaction data to iterate on spatial computing interface designs.
May 2026 - Present
Security
Cybersecurity Operations Center Student Analyst
Analyze and mitigate security incidents across the university’s network infrastructure, triaging potential threats and coordinating department-wide incident response.
Leveraged automated security orchestration tools to monitor enterprise-scale playbooks, identifying indicators of compromise (IOCs) and malicious URLs. Conducted forensic analysis of endpoints by cross-referencing MAC addresses, hostnames, and DHCP logs to identify compromised campus assets and isolate threats.
Responsibilities & Impact
- Investigated and triaged over 100+ phishing campaigns and user-reported security alerts weekly using Cherwell workflows.
- Monitored Palo Alto Cortex XSOAR playbooks to track outbound command-and-control (C2) traffic and cloud infrastructure anomalies.
- Coordinated with campus IT leads to initiate endpoint isolation and password reset procedures for compromised accounts.
Fall 2025 - Present
Research
Undergraduate Research Scholar
Conducted machine learning research on virtual sensing architectures for automotive power systems, resulting in an award-winning technical paper.
Formulated a unified deep learning model to estimate engine combustion states and NOx emissions. Collaborated with mechanical engineering faculty to preprocess real-world engine sensor datasets, design neural architectures, and validate predictions against classical regression baselines, culminating in winning the Central State Section of The Combustion Institute Undergrad Research Competition.
Responsibilities & Impact
- Preprocessed and normalized high-dimensional diesel engine sensor data spanning 217 operating points.
- Designed, trained, and optimized artificial neural network (ANN) architectures in Python and PyTorch.
- Authored and presented a technical paper at the Combustion Institute CSSI competition.
Jan 2026 - May 2026
Operations
L&S IT Service Desk Technician
Managed IT support and systems maintenance for the College of Letters & Science, resolving multi-tier technical issues for faculty, staff, and students.
Troubleshoot critical network, hardware, and OS issues across Windows and macOS environments. Maintained the infrastructure and software configurations of academic labs, implementing regular diagnostics to minimize downtime and ensure seamless instructional operations.
Responsibilities & Impact
- Resolved 30+ hardware and authentication service requests daily with high first-contact resolution rates, documented in Halo.
- Diagnosed and repaired instructional equipment, printers, and network connections in 3+ college computing labs.
- Provisioned and configured enterprise workstations, ensuring compliance with campus-wide security protocols.
Academics
Course Guide
Projects
Selected projects.
Led development of an AI-powered sign language learning and recognition platform combining computer vision, machine learning, and wearable AR technology to make sign language learning interactive and accessible.
The platform uses hand landmark detection and gesture recognition to identify American Sign Language (ASL) gestures in real time. Learners receive immediate visual feedback while practicing signs, creating a more engaging and accessible learning experience.
Key Features & Contributions
- Real-time hand tracking and keypoint isolation using MediaPipe.
- AI-powered ASL gesture classification models with optimized inference latency.
- Interactive educational interface offering real-time feedback on signing accuracy.
- Modular architecture designed to easily support future wearable AR integration.
- Formulated and implemented data processing, normalization, and model evaluation pipelines.
Designed and developed a multi-target virtual sensing framework using a single artificial neural network (ANN) to replace multiple expensive physical sensors in a 12L 6-cylinder diesel engine.
Built using a shared-representation deep learning architecture, the model learns latent representations to simultaneously predict heterogeneous, non-linear thermodynamic parameters (MF_IA, NOx_EO, SOC) in real time. The framework achieved high test accuracy (R² > 0.98 for all targets) across 217 engine operating points.
Key Features & Contributions
- Engineered multi-output ANN architectures (ANN-6 and ANN-13) in PyTorch for non-linear thermodynamic engine states.
- Preprocessed and analyzed real-world high-fidelity experimental data across full engine load and speed ranges.
- Achieved predictive R² scores of 0.9945 for intake air mass flow, 0.9891 for NOx emissions, and 0.9841 for start of combustion.
- Validated model generalization using stratified splits against Linear Regression, Random Forest, and SVR baselines.
July 18, 2026 · Participant
Industrial Asset Intelligence (Spatial Intelligence Ideathon)
Designed a spatial computing solution for industrial manufacturing environments during a one-day interdisciplinary startup ideathon, preserving equipment maintenance knowledge and reducing downtime.
Collaborated with a multidisciplinary team to commercialize Spatial Grounding and Awareness for Augmented Reality (SGA-AR) in industrial spaces, enabling spatial anchoring of machinery documentation directly onto physical assets.
Key Features & Contributions
- Formulated product strategy and technical architecture for shared AR spaces in manufacturing environments.
- Conducted rapid customer discovery and developed sustainable B2B SaaS business models.
- Led pitch deck creation and high-fidelity product dashboard demonstrations.
- Designed spatial anchoring logic to ground equipment documentation to physical machinery.
Summer 2026 - Present · Researcher
AR/VR Accessibility & Spatial HCI
Research project focusing on accessibility in immersive environments, utilizing depth-sensing cameras and spatial computing to assist low-vision users with real-world navigation.
Developed spatial intelligence prototypes that capture 3D environments and translate them into multi-modal cues (audio, haptics) for users. Investigated spatial layout grounding algorithms and user-testing protocols to ensure interfaces degrade gracefully under vision impairments.
Key Features & Contributions
- Leveraged Intel RealSense D455 depth data to map physical obstacles and estimate distances in real time.
- Formulated 3D spatial audio feedback models to communicate navigation markers and hazard warnings.
- Integrated gesture-based control paths to simplify VR interface navigation for accessibility.
Spring 2026 · Ongoing
Self-Refining Image Generation & Editing
Developed an iterative diffusion-based image generation framework featuring a closed-loop feedback mechanism for precise, localized image editing.
Built in PyTorch, the system utilizes latent representations and cross-attention maps to identify target editing regions. By integrating a multi-stage feedback loop, users can progressively refine generated images through natural language prompts, resolving typical drift issues in single-step generation.
Key Features & Contributions
- Implemented local cross-attention control to isolate edits without affecting background pixels.
- Formulated a closed-loop feedback pipeline that scores and refines image details iteratively.
- Optimized diffusion step sequencing, reducing inference times for image-to-image tasks.
Designed and built an AI-assisted meal planner and personalization engine during the CheeseHacks 2026 hackathon, leading a 4-person development team.
Integrated large language models (LLMs) with dietary databases to deliver customized nutritional recommendations based on user health parameters. Built backend using FastAPI and structured responsive frontend in React, creating an interactive portal that auto-generates grocery lists and meal plans.
Key Features & Contributions
- Integrated GPT-4o API for personalized, context-aware recipe recommendations.
- Developed a FastAPI backend with asynchronous query handling for rapid response times.
- Constructed a React-based frontend dashboard with visual recipe cards and nutritional breakdowns.
Built an interactive schedule optimization application at the Badger Build Festival to help university students build conflict-free course schedules.
Developed a search and filter system that runs constraint-satisfaction logic client-side in JavaScript. The tool parses course timings, prerequisites, and workload metrics, presenting students with a visual calendar dashboard and alternative schedule configurations.
Key Features & Contributions
- Built a client-side constraint solver to filter schedules based on user availability and load limits.
- Designed an interactive grid calendar showing overlapping slots and walking times between classes.
- Created a clean, responsive HTML/CSS dashboard with smooth drag-and-drop course selections.
Fall 2025
Human Survival Probability Analysis
Conducted a statistical and machine learning study analyzing socioeconomic factors and their correlations with human survival outcomes across ten large-scale datasets.
Performed rigorous data cleaning, missing value imputation, and exploratory data analysis using Pandas and Seaborn. Trained and evaluated multiple classification models, including Artificial Neural Networks (ANNs) and logistic baselines, comparing performance using area-under-the-curve (AUC) and receiver-operating-characteristic (ROC) metrics.
Key Features & Contributions
- Processed and merged ten heterogeneous datasets, handling class imbalances and multi-collinearity.
- Trained and optimized multi-layer ANN classifiers in Python.
- Produced publication-ready statistical visualizations of feature importances and correlation matrices.
Skills
Core tools and technical areas.
Programming
Python, Java, JavaScript, HTML5, CSS3
Frameworks & Tools
React, FastAPI, Git, GitHub, VS Code, PyTorch
Data & Modeling
Pandas, NumPy, Matplotlib, Seaborn, Scikit-learn, ANNs
Focus Areas
Cybersecurity, HCI, accessibility, computer architecture, AI tooling
Interests
Reading
Exploring technical books, research papers, and fiction
Cinema
Watching movies ranging from classics to horror films
Music
Listening to diverse genres and discovering new artists
Museums
Visiting art galleries and cultural exhibitions
Favorite Bookshelf
Hover over or tap a book spine to reveal insights.
Nexus
1984
WSJ
Thinking Fast/Slow
Why Nations Fail
Hover or tap on a book above to inspect details.
Cinematic Film Track
Hover or tap a frame to view filmmaker details.
I
Parasite
Bong Joon-ho (2019)
II
The Shining
Stanley Kubrick (1980)
III
Spirited Away
Hayao Miyazaki (2001)
IV
The Bad Seed
Mervyn LeRoy (1956)
Sound Waves
Hover or tap the translucent bubbles to reveal popular tracks.
Starboy
The Weeknd
Blinding Lights
The Weeknd
Bohemian Rhapsody
Queen
Perfect
Ed Sheeran
Cultural Exhibits
Visiting historical galleries, computing exhibits, and classical museums.
Art & History Galleries
Appreciating neoclassical oil paintings, neoclassical sculptures, and historical artifacts.
Science & Tech Exhibits
Fascinated by mechanical computation history, structural physics setups, and structural engineering marvels.
Contact
Let’s connect.
For research, engineering, internships, or collaboration, you can reach me here.
