- Deployed LLM/LMM models for POC projects on-prem and cloud.- Developed and implemented a multi-node AI model training pipeline to increase the max batch size of the model training process.- Implemented model and dataset versioning systems to ensure traceability and reproducibility of experiments.- Designed and implemented active learning strategies for large-scale data to reduce annotation costs and improve model performance.- Deployed real-time containerized auto-scaling ML services in the cloud to enable elastic and cost-efficient scaling of machine learning applications.- Built a real-time service monitoring system to detect and diagnose issues in production environments and improve system reliability.

- Implemented state-of-the-art semi-supervised instance segmentation models to improve accuracy and reduce labeling costs.- Designed and implemented multi-modal CLIP models, combining text and image data, to enable real-time zero-shot classification.

[Champion] Contributor to Lightning AI - Pytorch Lightning

Part of a selected group of 50 developers to test and contribute to features before public launch.

- Conducted data analysis program for Distribution Team, resulting in improved visibility and understanding of supply chain data.- Developed time-series machine learning prediction models for product distribution across several warehouses around all Korea, which reduced distribution costs and improved delivery times.- Collaborated with other teams to identify opportunities for utilizing big data and machine learning in business processes.

Published 10+ articles on Yonsei University Research Center's newsletterResearched about:- Privacy and security- Artificial Intelligence- New ICT technologies

Skin cancer image classification model- Presented the skin cancer classification model in a poster presentation at the Smart HealthCare Conference in Seoul, raising awareness about the potential of AI in healthcare and sharing the results of the project with a wider audience.- Deployed an API for the skin cancer classification model via a web application, enabling easy and user-friendly access to the model for healthcare professionals and patients.- Collaborated with senior members of the team to identify opportunities for further applying AI in healthcare and contributed to the development of innovative healthcare solutions.

Transform raw big data about electricity usage of different apartments/districts in Seoul into readable data and perform several statistical analyses/visualizations.

- Korean <-> Portuguese- Korean <-> English