Kairui Zhu / kerrypig

Kairui Zhu.

Statistics. Intelligence. Things that work.

University of Toronto undergraduate studying computer science, statistics, and applied mathematics. I build RAG pipelines, LLM applications, and interactive tools.

Explore projects Get in touch

University of TorontoUofT ’27

ALWAYS CURIOUS. ALWAYS BUILDING.SCROLL TO EXPLORE

01 / About

From evidence to working software.

My work connects statistical thinking with practical AI engineering. I have contributed to vehicle-manual question answering, evidence-traceable training data, and a multimodal CAD plugin prototype. I’m interested in how retrieval, models, and useful interfaces fit together.

Current interests

RAG & evidence-grounded answersLLM applications & agent workflowsStatistical learning & interactive tools

02 / Selected work

Ideas, built and explored.

A selection of AI engineering, machine learning, and statistical computing projects.

RAG & LLM applications · Jun–Jul 2026

Vehicle Manual RAG

A question-answering and training-data pipeline grounded in vehicle manuals. Combines FAISS and BM25 retrieval, query rewriting, vehicle metadata, and evidence checks with multi-turn context and structured dataset output.

Python
FAISS
BM25
Qwen
Ollama

Traceable evidence and JSON/JSONL output for SFT/LoRA data workflows.

View Vehicle Manual RAG source

Multimodal CAD prototype · Jun–Aug 2025

CrownCAD AI Plugin

A CAD plugin prototype connecting text, drawings, multiple views, and point-cloud inputs to model preview and export. Contributed to frontend/backend integration, upload and export APIs, and standard-part matching.

Vue
Element UI
Three.js
FastAPI

Team project · First Prize, CAD/CG & GDC 2025 Intelligent CAD Generation and Parametric Modeling Challenge.

View CrownCAD plugin source

Machine learning · Team coursework

Painting Classification

A course project predicting which painting a survey response describes. Combines structured features and TF-IDF text with logistic regression, with inference implemented using NumPy and Pandas.

Python
NumPy
Pandas
scikit-learn

90.24% hold-out test accuracy · 88.37% ± 0.83% in 5-fold cross-validation.

View painting classification source

Statistical computing · Team project

Bias–Variance Explorer

Contributed to an R package and Shiny app exploring model complexity through Monte Carlo simulation. Interactive charts compare regression models, decompose prediction error, and let users export the results.

R
Shiny
Plotly
Monte Carlo

View Bias–Variance Explorer source

More projects

Java Paint · JavaFX drawing tool mysh · A shell in C

03 / Experience

Learning through real problems.

May–Jul 2026

Huawei

LLM Algorithm Intern

Contributed to vehicle-manual RAG and evidence-traceable training data for in-car assistants. Worked on vehicle-specific retrieval, evidence verification, and structured JSON/JSONL output for SFT/LoRA workflows.

May–Aug 2025

Harbin Engineering University

Research Assistant

Contributed to multimodal CAD research and a CrownCAD plugin prototype, connecting data preparation, model previews, standard-part matching, and export workflows. Helped prepare more than 300 multimodal standard-part samples.

04 / Education

A foundation in data and computation.

Sep 2023–Jun 2027 · Expected graduation

University of Toronto

Honours Bachelor of Science · In progress

Computer Science major · Statistics · Applied Mathematics

GPA 3.7 / 4.0Dean’s List

Machine learningDeep learningTime series analysisStatistical learningHigh-dimensional data methodsDynamical systems and chaos

05 / Contact

Open to AI / RAG internships

Let’s build something useful.

I’m interested in AI, RAG, and LLM application internships, as well as opportunities to collaborate on thoughtful software.

kairui.zhu@mail.utoronto.caFind me on GitHub