Jaekang Lee’s AI blog – jaekangai-quarto

How Close is ChatGPT to Brain? (Engineering Point of View)

ai

neuroscience

AGI

A neuroscience-informed analysis of what’s missing in current AI systems like ChatGPT compared to human brain architecture, and the concrete engineering roadmap to AGI.

Agentic Text-to-SQL System for Enterprise Databases

ml

RAG

agent

A case study documenting the development of Woody, an enterprise text-to-SQL system that combines four specialized RAG components with agentic workflows to handle complex business queries across 81 database tables in the lumber industry.

Audio RAG: Segment-Based Music Recommendation System

ml

RAG

audio

music

recommendation

prototype

lessons-learned

This project began with a simple observation: we fall in love with just a few seconds of a song, not the entire song. The journey led to building a system that could search for those moments, not just whole tracks. Along the way, I discovered the practical challenges of audio data, the limits of small-scale experiments, and the real difference between a prototype and a product.

Advanced Machine Learning Optimization Techniques: From Hyperparameter Tuning to Model Fine-tuning

ml

optimization

hyperparameters

deep-learning

kaggle

A comprehensive guide to modern ML optimization techniques including Bayesian optimization with TPE, hyperparameter tuning strategies, and efficient model fine-tuning methods. In other words, machine learning techniques that I randomly stumbled upon andfound fascinating on Kaggle.

Hybrid Search and Reranking in RAG: When More Sophisticated ≠ Better Results

ml

RAG

hybrid-search

Comprehensive evaluation of hybrid search strategies and reranking methods in RAG systems, revealing surprising insights about when simplicity outperforms complexity.

Contextual Chunking vs Fixed-Size: A Comprehensive RAG Performance Analysis

ml

RAG

chunking

Empirical evaluation of Anthropic’s contextual chunking strategy against traditional fixed-size approaches, revealing optimal chunk sizes and the impact of context augmentation on retrieval accuracy.

RAG Evaluation at Scale: A True/False Methodology for Multilingual Systems

ml

RAG

evaluation

A scalable approach to evaluating RAG pipeline accuracy using binary fact verification, with experimental results showing multilingual embedding improvements.

Kaggle Sept, Oct 2024

kaggle

EDA

ml

Cancer ISIC detection 2024 + Ariel Exoplanet data 2024 challenge

Kaggle August 2024

kaggle

EDA

ml

Binary classification using XGBoostClassifier + Predicting human processing behavior when humans are reading AI outputs using llama3.1.

Book blog 2024 V1 📚

book

12rules how and why to rich habit at 5am making bed quietly with strangers.

CV generator using LangGraph agentic workflow (team of AI’s) 🤖

ml

side-quest

note: may take 50 second to load as it is free tier.

Non-medical Drug Use in Canada 🍁

EDA

datafest

DataFest 2021 @ University of Toronto

Fairness in Hiring and Salary Statistical Report ⚖️

visualisation

EDA

An analysis of fictional company Black Saber

Jane Street Market Prediction 🎯

Kaggle

ml

Jane Street Market Prediction Kaggle Competition

Jane Street Market EDA 📈

EDA

Kaggle

Jane Street Market Prediction Kaggle Competition

Recommendations with IBM 😺

recommendation system

Udacity

Udacity project with IBM Watson Studio platform data

My friend’s bike data interactive visualisation🚴

visualisation

Building interactive webapp with Python, Bootstrap and Flask

MOOC certificates 📜

certificate

LoL Prediction S11 🗡️

nlp

side-quest

League of Legends s11 Ranked Prediction

Deep Art Gallery 🧠

computer_vision

side-quest

Neural Style Transfer, ConvNet(VGG19), Transfer Learning, tf-gpu

So I learned some new algorithms.. 🃏

meme

Learning with memes

LoL Prediction S10 🏹

nlp

side-quest

LOL s10 high elo ranked games prediction.

City Detector 🏙️

computer_vision

side-quest

Coquitlam Paris Seoul and New York.

Crocodile plush detector 🐊

computer_vision

side-quest

Along with my hand drawn arts.