Building AI systems that reason over knowledge.
python · rag · langchain · nlp · RL · vector-db · ollama
I'm Thura Win Kyaw, an AI/ML Engineer based in Daejeon, South Korea. I design and build RAG pipelines, LLM applications, and conversational AI systems — currently developing intelligent biomedical knowledge retrieval at AIdenBio. My work sits at the intersection of large language models, retrieval systems, and real-world software, turning research ideas into production-grade AI applications.
I specialize in Retrieval-Augmented Generation (RAG), LangChain-based pipelines, and LLM-powered conversational systems. I enjoy designing end-to-end AI workflows — from vector search and prompt engineering to FastAPI backends and local model deployment with Ollama.
Alongside my engineering work, I'm completing a Bachelor's degree in Electrical Engineering at Chungnam National University, where I explore the overlap between control systems, signal processing, and intelligent automation. I hold HarvardX certifications in Computer Science, Python, and AI.
Research Interests
Protein Structure Visualization Assistant
An intelligent assistant that interprets natural language queries to visualize 3D protein structures from the RCSB PDB. Integrates LLMs with a RAG pipeline for semantic understanding of protein documentation, and PyMOL for interactive rendering — enabling biomedical researchers to query structural data conversationally without writing visualization scripts.
System Architecture
Python · SVM · Scikit-learn · PCA
Supervised SVM classifier with StandardScaler preprocessing and GridSearchCV hyperparameter tuning. Evaluated via accuracy, confusion matrix, and F1 score. Decision boundaries visualized with PCA; best model pipeline serialized for reuse.
Python · K-means · Scikit-learn
Unsupervised K-means clustering on customer demographic data to identify actionable segments. Elbow method and silhouette analysis for optimal k selection. Trained model and scaler serialized for downstream deployment.
Python · NLP · TF-IDF · SVM · Logistic Regression
Multi-model text classifier with TF-IDF features across Naive Bayes, Logistic Regression, and SVM with GridSearchCV tuning. Evaluated with precision, recall, and F1. Model, vectorizer, and label encoder saved for deployment.
Python · NLP · VADER · BeautifulSoup
End-to-end sentiment pipeline for short quotes — scraped with requests and BeautifulSoup, labeled with VADER, trained with Naive Bayes and Logistic Regression. Evaluated with confusion matrices and word cloud visualizations.
AIdenBio · Daejeon, KR
GRINDA AI · Daejeon, KR
Built an AI-powered Slack bot that automatically converts issue reports into structured GitHub issues using Claude AI and FastAPI, with auto-labeling, translation, and monitoring.
Chungnam National University · Daejeon, KR
Relevant coursework: Computer Programming (C), Linear Algebra, AI and Future Society, Modern Control Theory and Design, Sensor and Measurement Engineering.
