Giwon Baek

News

Education

• 

  Sogang University, B.S. in Computer Science and Engineering

  Mar. 2020 - Feb. 2027

  • Relevant Coursework: Computer Vision, Basic Machine Learning, Deep Learning, Introduction to Artificial Intelligence, Operating Systems

  • Research Interests: Dexterous Manipulation, Affordance-Aware Manipulation, Cross-Embodiment Learning, 3D Vision, Vision-Language-Action (VLA) Models

Experiences

• 

  Korea Institute of Science and Technology (KIST) HARI Lab

  Dec. 2025 - Jun. 2026

  Robotics Research Intern

  • Developed affordance-aware dexterous grasp generation pipelines from 3D object point clouds using vision-language models and part-aware segmentation.

  • Applied a cross-embodiment model to transfer dexterous manipulation policies to a KISTAR hand, a custom robotic hand developed at KIST.

  • Collected humanoid teleoperation data for dexterous manipulation tasks using VR-based teleoperation systems.

  • Constructed dexterous grasp pose datasets for the KISTAR hand using retargeting methods, open-source models, and Isaac Gym-based validation.

• 

  Like Lion AI CV Bootcamp

  Sept. 2025 - Dec. 2025

  Computer Vision Bootcamp Participant

  • Improved RT-DETR-based object detection for autonomous driving scenarios using a Vision Mamba encoder.

  • Enhanced medical image captioning by integrating segmentation masks with RGB breast tissue images.

  • Developed a real-time deepfake detection video call application; received first place in the KEIT AI Life Solution Challenge.

• 

  Doosan Robotics ROKEY Bootcamp

  Mar. 2025 - Sept. 2025

  Robotics Software Bootcamp Participant

  • Completed intensive training in ROS 2, computer vision, and autonomous navigation.

  • Developed a lane-following autonomous driving system for TurtleBot3 using a downward-facing camera.

  • Built a full autonomous mobile robot pipeline using TurtleBot4, RGB-D, LiDAR, YOLO, SLAM, and Nav2.

  • Developed a voice-guided kitchen assistant robot using YOLO segmentation, PCA-based grasp estimation, and MediaPipe gesture interaction.

• 

  Learning Lab Co., Ltd.

  Jun. 2023 - Jul. 2023 & Jan. 2024 - Feb. 2024

  Software Engineer Intern

  • Participated in government-funded smart manufacturing and Data Voucher projects.

  • Developed software pipelines for real-time manufacturing KPI collection and deployed sensor-based production monitoring systems.

  • Collected and annotated segmentation datasets for sign language recognition and automotive defect detection tasks.

Projects

• 

  Tactile Intelligence in Robotic Hands — KIST HARI Lab

  GitHub 2026

  • Developed an affordance-aware dexterous grasp generation pipeline from segmented 3D point clouds.

  • Integrated vision-language models, part-aware segmentation, and grasp pose generation.

  • Generated training data using NVIDIA Isaac Gym and deployed on a real robot via ROS 2.

• 

  Graspium

  GitHub 2026

  • Independent research seminar on dexterous manipulation; weekly paper discussions.

  • Topics included dexterous grasping, multimodal VLA, reinforcement learning, and imitation learning.

  • Presented papers including AffordGrasp, D(R,O), and UniMorphGrasp.

• Like Lion AI CV Bootcamp — 3 Sub-projects

  Sept. 2025 - Dec. 2025
  RT-DETR 자율주행 Object Detection

  

  GitHub

  • Vision Mamba encoder를 적용해 RT-DETR 기반 자율주행 객체 탐지 성능 향상

  • Technology: PyTorch, RT-DETR, Vision Mamba

  의료 이미지 Captioning

  

  GitHub

  • 분할 마스크와 RGB 유방 조직 이미지를 결합해 의료 이미지 캡셔닝 성능 향상

  • Technology: PyTorch, Segmentation, Transformer

  딥페이크 탐지 영상통화 앱 — KEIT AI Challenge 1위

  

  GitHub

  • CNN 기반 실시간 딥페이크 탐지 기능이 내장된 Flutter 영상통화 앱 개발

  • SigLIP 서버 분석 및 Grad-CAM 시각화로 설명 가능한 탐지 리포트 생성

  • Technology: Flutter, CNN, SigLIP, Grad-CAM

• Doosan Robotics ROKEY Bootcamp — 4 Sub-projects

  Mar. 2025 - Sept. 2025
  TurtleBot3 차선 추종 자율주행

  

  GitHub

  • 하방 카메라를 이용한 TurtleBot3 차선 추종 자율주행 시스템 개발

  • Gazebo 시뮬레이션에서 실제 로봇까지 Digital Twin 배포 경험

  • Technology: ROS 2, OpenCV, TurtleBot3, Gazebo

  TurtleBot4 자율 이동 로봇

  

  GitHub

  • RGB-D 카메라, LiDAR, YOLO, SLAM, Nav2를 결합한 완전 자율 이동 로봇 파이프라인 구현

  • Technology: ROS 2, TurtleBot4, YOLO, SLAM, Nav2

  삼겹살 그릴링 로봇 서비스

  

  GitHub

  • 삼겹살을 자동으로 굽는 로봇 서비스 시스템 설계 및 구현

  • Technology: ROS 2, Computer Vision

  음성 안내 주방 보조 로봇

  

  GitHub

  • YOLO segmentation, PCA 기반 파지 방향 추정, MediaPipe 손 제스처 인식을 결합한 주방 보조 로봇 개발

  • Technology: ROS 2, YOLO, PCA, MediaPipe

• 

  KEIT AI Life Solution Challenge — 1st Place

  2025

  • Developed a Flutter-based video call app with CNN-based real-time deepfake detection.

  • Performed server-side SigLIP analysis on suspicious frames and generated Grad-CAM explainability reports.

  • Award: First Place — Chairman’s Award of the KEIT Evaluation Committee

• 

  Hecto AI Challenge — 2nd Place

  2026

  • Developed a deepfake detection model based on DINOv2.

  • Constructed a large-scale dataset of 0.6 million images for training and preprocessing.

• 

  LG Aimers 5th — Top 6%

  2024

  • Developed an AI model to detect defective products using sensor data from LG Display manufacturing processes.

  • Ranked in the top 6% among more than 750 participating teams.

Awards

• 1st Place — KEIT AI Life Solution Challenge

  2025

  Chairman’s Award of the KEIT Evaluation Committee (Like Lion AI CV Bootcamp)

• 2nd Place — Hecto AI Challenge

  2026

  Deepfake detection model based on DINOv2, dataset of 0.6M images

• Top 6% — LG Aimers 5th

  2024

  AI defect detection using LG Display manufacturing sensor data (750+ teams)

Skills

• Programming Languages

  

• Frameworks & Tools

  

Languages

• Korean (Native): ●●●●●

• English (Fluent, OPIC AL): ●●●●○

• Portuguese (Intermediate): ●●●○○