Hi, I'm Hyeontae Kim.

Data Science Co-op @ U of Waterloo


I’m passionate about neural networks and love exploring how they can understand world better. I enjoy and try diving into research papers to stay on top of the fast-paced AI field and applying cutting-edge techniques to real-world challenges. I’m also keen on data engineering —it’s the backbone of all things data, after all. Looking ahead, I see myself as a multimodal AI Agent engineer, working on systems that seamlessly combine vision, language, and more. Fingers crossed!

Julia Walker

Latest Project

Picnic
Visit Repository

Deep Learning from Scratch (CV, NLP)

Hands-on exploration of deep learning, spanning classical CNNs like AlexNet and VGG to cutting-edge transformers like BERT and GPT. Written entirely in Python with PyTorch, this demonstrates my ability to pipeline diverse architectures and downstream tasks in computer vision (classification, detection, segmentation) and NLP (translation, NER, next-token prediction). Each model is built “from scratch” to reveal core architectures of each valuable milestone in modern AI.

Picnic
Visit Repository

TechCrunch Q&A with RAG-Powered LangChain

Simple question-and-answer system specialized for TechCrunch Articles using LangChain, OpenAI embeddings, and Chroma for vector storage. Load and segments TechCrunch articles and converts them into embeddings that are persisted for efficient retrieval. Most relevant document chunks are retrieved from Chroma and passed to an OpenAI GPT-like LLM for final answer synthesis, ensuring reduced hallucinations by grounding responses in source documents.

Picnic
Visit Repository

AI arXiv Daily (High-quality Translator into Korean)

A fully automated pipeline that fetches, filters, and translates AI-related research papers from arXiv into Korean on a daily basis with AirFlow scheduling. Leveraging OpenAI’s GPT API with advanced prompt engineering, the script ensures context-aware translations. Translated papers are compiled into high-quality PDF documents, enhancing accessibility for the Korean research community.

Picnic

Building a C Language Compiler from Scratch

Built a compiler from scratch for a subset of the C language using C++. The compiler meticulously transforms high-level C code into binary code, encapsulating all necessary stages of the compilation process. This includes scanning, parsing, context-sensitive analysis, and code generation using algorithms like Simplified Maximal Munch and Bottom-up parsing.

My Knowledge Hub in Deep Learning / Machine Learning

I read research papers, take notes, share ideas via presentations, and implement codes in PyTorch.
(Few of them not ready just yet!)

Computer Vision

NLP

Machine Learning

ML Essentials