Hyunwook Jo

Project

2025 Indoor Accessibility Analysis AI System for the Mobility-Impaired (2025.07 ~); I am currently leading the development of an indoor accessibility analysis AI system for mobility-impaired individuals, with the support of Kakao Impact, Tech for Impact, and Dongjeop Lab. To address the critical lack of accessibility information, I have initiated a vision AI system that can analyze indoor spaces from a single image. This involves applying a multimodal LLM based on the Gemma-3-4B-IT model for training, designing an efficient deployment environment with the vLLM engine, and establishing a dual-pipeline architecture using AWS S3 and MySQL for stable data management. Beyond the technical aspects, this project focuses on creating social value by actively collaborating with the "Stair Buster Club" campaign to incorporate the voices of the mobility-impaired, aiming to foster social empathy and voluntary participation. We are also in the process of building and preparing to distribute an image captioning dataset for this community to be utilized by a wider audience of researchers and developers.
2025 Incheon International Airport Corporation PoC Project (2025.03 ~); Development of Meeting Room Reservation System and Robot Demonstration;
Led the development of a complex AI-based PoC for a real client organization.
For the meeting room reservation system, designed natural language-based real-time optimized search and CRUD API automation features, and implemented SQL generation and external vendor API integration via MCP Agent.
For the robot demonstration (Physical AI), designed MCP Agent to interpret user commands and execute control instructions such as coffee preparation and delivery.
Defined and tested the robot communication flow, including ROS-based path planning, sensor integration, and external hardware interfacing.
The project aims to secure demonstration materials for Ministry of Science and ICT stage evaluation and explore the feasibility of transitioning to an internal production system.
2024 ART KOREA LAB ARTxTECH Research Labs (2024.10 - 2024.11); Proposed and led the development of an AI-based online portfolio system for young artists; Developed an Invisible Watermark model; Enhanced watermark detection performance by training YOLO models; Optimized AI model deployment by converting to ONNX for CPU inference to reduce AWS EC2 costs; Organized and led a team to develop a user-friendly web platform using React and Spring; Won the Grand Prize at the Inha University Startup Competition (October 2024) and the Excellence Award at the ART KOREA LAB Hackathon (November 2024).
2023 ARTS KOREA LAB ARTxTECH Research Labs (2023.07 - 2023.11); Led AI Video research team, studying CS231N and Coursera courses on generative models; Developed an audio-driven denoising intensity control technique for Stable Diffusion to enhance AI-generated music videos; Built a React-based interactive demo booth; Presented research at MODUCON 2023 AI Conference.
2023 Intelligent Information SW Academy, Korea University (2023.03 - 2023.06); Developed AI Math Notes for Hankyung (Korea Economic Daily); Conducted the Senty Project, training T5 and BERT models on sentiment analysis datasets, performing topic modeling, crawling Twitter, retrieving S\&P 500 company data, performing sentiment analysis, and computing Pearson correlations.
2022 AI NLP Expert Training Course, KAIST-supported K-DIGITAL TRAINING Boot Camp (2022.07.08 - 2022.11.04); Studied Python to Deep Learning (DNN, CNN, RNN, GAN, Transformer); Participated in Kaggle 5th Goorm projects: Text Classification using BERT (2022.09.27 – 2022.10.04) and Korean MRC using T5 (2022.10.05 – 2022.10.18), responsible for Data Preprocessing, Modeling, Ensemble, and Visualization.
2022 Capstone Design, Incheon University (2022.09 - 2022.12); Studied data collection, preprocessing, Transformer model architecture, and Gradio; Developed a Machine Translator Project as a one-person team, responsible for Data Collection, Preprocessing, Modeling, and Service Implementation.

Similar Profiles