HELLO, I'M

Chenyu Wei.

Computer Science student.
AI & systems builder.

From multimodal AI to database internals,
I turn complex ideas into working software.

Explore projects Download CV

University of Waterloo

GET TO KNOW ME

01 / THE PERSON BEHIND THE CODE

About me.

Curious about systems.
Thoughtful about people.

I'm a Computer Science student at the University of Waterloo, pursuing a Bachelor of Mathematics with a Human-Computer Interaction specialization.

My work spans multimodal AI pipelines, database research, and practical developer tools. I enjoy connecting the pieces: data, models, infrastructure, and the people who use them.

Read my full resume ↗

Applied experience

AI application development
Database systems research

TWO INDUSTRY INTERNSHIPS

University of Waterloo

BMath · Computer Science
Human-Computer Interaction

SEP 2024 — JUL 2029

Languages

Python · C++ · Java
JavaScript · SQL · Bash

AI & data

PyTorch · Transformers · Unsloth
Ollama · OpenCV · pandas

Systems & databases

Linux · Docker · GitHub Actions · Nginx · Git
PostgreSQL · MySQL · SQLite · DuckDB

02 / LEARNING BY BUILDING

Experience.

MAY — AUG 2026

Dameng Database
Company

Onsite internship

R&D Research Intern

Reproduced an LLM-based agent for database-native function synthesis across SQLite, PostgreSQL, and DuckDB.
Automated generation, compilation, testing, packaging, and deployment for DBCooker and SemanticCE research implementations; built model training and fine-tuning experiment pipelines.
Co-developed a Dify and MCP workflow to classify and summarize 378 database papers into structured JSON.
Built a bilingual paper translation skill preserving images and LaTeX formulas. Published three technology weeklies and presented results to the research department.

LLM agentsDatabase systemsMCP

MAY — AUG 2025

Black Sesame
Technologies

Onsite internship

AI Application Intern

Developed multimodal AI pipeline components for AEB decision support using 50,000+ road images, radar-derived features, and vehicle-state signals.
Reduced visual input frames by approximately 80% through key-frame selection, with schema validation, retries, and resumable processing for structured data.
Implemented scenario classification and contributed to team-based Qwen-VL LoRA fine-tuning, monitoring, and offline evaluation.

PyTorchQwen-VLOpenCVDocker

03 / SELECTED WORK

Projects.

Biquadris

A two-player, Tetris-like C++ game built with classmates, with modular game logic, extensible rules, and text and graphical interfaces.

C++OOPUML

View repository

02 / INFRASTRUCTURE

ShellOps

Idempotent Ubuntu provisioning and a lightweight Bash CI pipeline. Reduced server setup from about two hours to twenty minutes.

BashLinuxCI/CD

View repository

03 / RESEARCH TOOLING

paper-duiyi

A PDF-to-bilingual-Markdown pipeline that parses papers, checks formulas, and aligns English and Chinese while preserving images and LaTeX.

PythonMinerULaTeX

View repository

04 / DATABASE INTERNALS

cydb

A small database in C exploring persistent storage, B+ tree indexing, ordered record retrieval, and the mechanics behind a database engine.

CB+ treeStorage

View repository

05 / EVERYDAY AUTOMATION

US Visa Rescheduler

An appointment availability monitor with Python and browser-extension options. Sends notifications while keeping booking decisions in the user's hands.

PythonPlaywrightBrowser extension

View repository

“Ask a question.”↓

SELECT answers
FROM your_data;

06 / AI & DATABASES

Text2SQL Agent

A three-stage pipeline for schema retrieval, SQL generation, and database-backed validation, with bounded self-correction and inspectable execution traces.

PythonSQLAgents

View repository

04 / IDEAS & EXPLORATIONS