At RealityMine, I developed two PySpark e-commerce web-shopper feeds on AWS and took them to production. I transformed raw web traffic across macOS, Windows, iOS, and Android into structured behavioural datasets for downstream analytics.
I delivered the feeds through Bitbucket pull requests and Jenkins CI/CD, with jobs running on AWS EMR Serverless through Airflow. I also wrote integration tests and validation checks, then used Athena queries to validate staging runs.
At Deltatech Korea (BRM / Delbiznet), I developed a desktop semantic-search application using LangChain, FAISS, HuggingFace embeddings, and Llama 3 via Ollama. I also built a Pandas data-cleaning pipeline that auto-tagged entries by industry sector and geography.
For my final-year project, EmotionalTranquilityAI, I built a multimodal emotion-recognition system with custom CNN and CNN-LSTM models trained from scratch. I also built an offline AI document-search tool that answers PDF questions with source passages, filenames, and page references.

