Professional index · Korea · Updated October 2026 · Korean

Minsoo Kim

AI Research Engineer · Samsung Electronics

Minsoo Kim during the AI Vision Inside developer interview
AI Vision Inside developer interview · Watch video ↗
Mobile vision and 3D modeling development
Mobile Vision
Smart TV content recognition demonstration
Smart TV
AI Vision Inside product presentation
AI Vision Inside

  1. Mobile Vision

    2012–2016

    Camera and gallery vision
    3D face and object modeling

  2. Vision Recognition
    Recommendation System

    2016–2022

    Smart TV content recognition
    GAN photo expansion · Graph recommendation

  3. AI Vision Inside

    2023–2025

    Food recognition · Real-time NPU inference · Product delivery

  4. Post-Training (Agentic AI, Physical AI)

    2026–present

    Synthetic data · Data selection and refinement
    Dynamic data valuation

LinkedIn ↗ GitHub ↗ Google Scholar ↗ Email ↗ Work email ↗

Current Research

Samsung Research · Data Intelligence Team · 2026–present

Related publications · Related talks

I work on post-training in which a model generates, verifies and selects the data it learns from next. Within this loop of generation, verification, selection and training (Self-Evolving Data), I focus on selection: deciding which data to train on, and when.

  • Gain and cost of data selectionIn reinforcement learning with verifiable rewards (RLVR), I compare the training gain that data selection brings with the compute that selection costs, on a common scale.
  • Switching selection strategies during trainingThe value of data changes as training progresses. I study how to change the selection strategy partway through training to follow that change.
  • Data for agentic and physical AII am extending the same loop to the data used to train agents and physical AI systems.

Research & Product Delivery

Experience

Newest first · 15 roles and projects from the source record

  1. 2010.01–2012.06 Samsung Software Membership, Gangnam Software R&D and pre-employment projects
  2. 2010.06–2010.08 SAIT, 3D Mixed Reality Group Shader-programming-based high-fidelity 3D ultrasound visualization
  3. 2011 Software Maestro Mentee · co-translated and compiled a pattern language for study groups
  4. 2012.01–2012.04 Samsung HR Development Center Digit recognition via image processing and MLP; UK inspection automation software
  5. 2012.04–2012.07 Samsung DMC R&D, S#Project Team Galaxy Note handwriting recognition and Android applications
  6. 2012.07–2014.01 Visual Effect / Vision Core (MX·VD) Visual Effect Engine, Vision Core Engine, and apps for Android smartphones
  7. 2014.01–2016.03 Galaxy Gallery Apps/Service 3D face/object modeling, 360° panorama image algorithm
  8. 2016.03–2019.03 Smart TV Content Recognition Multilingual/mixed-language text and keyword recognition, TV–mobile content recognition, OSD-based automatic content recognition
  9. 2019.03–2020.01 Wi-Fi CSI Analysis Indoor user location and activity AI from time-series analysis
  10. 2020.01–2021.02 Samsung C-Lab Project Tech Lead · GAN-based image outpainting and vision-AI photo expansion for Galaxy Gallery
  11. 2021.02–2021.12 Session-Based Recommendation Tech Lead · SR-GCL graph-contrastive session recommendation — paper and patents
  12. 2023.01–2023.12 AI Vision Inside Tech Lead · core on-device food recognition for refrigerators, productized
  13. 2023.12–2024.12 AI Vision Inside Project Lead · designed 9 on-device AI models running real-time on NPU
  14. 2024.12–2025.12 AI Vision Inside 2.0 / Family Hub 2025 · Food Recognition Service Project Lead · food recognition, RL promotions, and feature matching; expanded to 37 fresh-food classes and up to 50 packaged foods; Instacart integration
  15. 2026.01–present Samsung Research Data Intelligence Team Post-Training (Agentic AI, Physical AI) · Data-Centric AI · Dynamic Data Valuation Research

Awards

Newest first · 5 individual and company awards · 6 product award contributions

  1. Samsung Super Rookie Project Company-wide selection · ranked 1st among new hires
  2. Samsung Research CTO Awards Top project of the year at Samsung Research
  3. Samsung Work Smart Contest — Encouragement Prize CFO Awards · project with outstanding cross-team collaboration
  4. CES 2024 Innovation Award Product award contribution — AI Vision Inside CES ↗
  5. Samsung Annual Awards — Gold Lead Engineer, AI Vision Inside
  6. Samsung SW Awards — Grand Prize (CEO Award) Ranked 1st company-wide · principal awardee
  7. Korea Innovation Awards — Convergence Innovation Product award contribution KSA ↗
  8. IDEA 2024 Product award contribution
  9. CES 2025 Innovation Awards Product award contribution — AI Vision Inside 2.0 / Family Hub 2025
  10. Korea Innovation Awards — New Technology Innovation Product award contribution Article ↗
  11. CES 2026 Innovation Awards Core food-recognition features, led Gemini-integrated service, wrote the CES entry

Patents

29 granted · 11 inventions · 81 filings across countries and routes · latest grant first

All patents on Google Scholar ↗ · open an invention to see its grants by country; each number opens on Google Patents ↗

  1. GNN-based viewer-group recommendation Electronic apparatus and controlling method thereof · US 1 grant

    Analyzes each user's viewing history with a GNN to derive content attributes and recommends content to groups with similar viewing.

  2. Fingerprint database for content recognition Electronic apparatus, control method thereof and electronic system · US · KR · EP · CN 4 grants

    Identifies objects in multimedia frames and sends content information and fingerprints to a server to build the recognition database.

  3. Broadcast channel-change detection Electronic apparatus and controlling method thereof · US · KR · EP · CN 4 grants

    Checks audio continuity around silent gaps in broadcast content to tell whether the channel changed.

  4. Object-based content identification Electronic apparatus for identifying content based on an object included in the content and control method thereof · US · KR · CN 4 grants

    Identifies objects in multimedia data from an external device and identifies the content using the server's content guide information.

  5. Content information on mobile devices Method and device for providing information on content · US · KR 2 grants

    A mobile device detects a nearby device showing content, requests object recognition, and provides related information.

  6. Content-change recognition Display apparatus and control method thereof · US · KR · EP 3 grants

    On a content-change event, extracts the UI pattern and fingerprint, sends them to a server, and receives the content information.

  7. Multilingual and mixed-language text recognition Method and terminal for recognizing text · US · KR · EP 3 grants

    Recognizes multilingual and mixed-language input from the input UI's language tags, cursor position, and a screen image.

  8. Vision-AI image outpainting Electronic device and controlling method of electronic device · US 1 grant

    Extends image borders naturally using object segmentation, then converts the result into an RGB image.

  9. Scene descriptions for multimedia Electronic apparatus and method for controlling the electronic apparatus · US · KR 2 grants

    Selects frames that contain relevant objects and generates descriptions for scene analysis.

  10. Wi-Fi CSI signal calibration Electronic device and control method thereof · US 1 grant

    Calibrates signals received from other devices with per-device correction data to identify the transmitting device.

  11. Text- and fingerprint-based content recognition Electronic device and content recognition information acquisition therefor · US · KR · EP · CN 4 grants

    Captures the content screen and obtains recognition information from text when present, otherwise from a fingerprint.

Published applications under examination · 6 selected
  1. US20260104197A1 Refrigerator and controlling method thereof Captures items in and out, guides additional images, stores food data Google Patents ↗
  2. KR20250154875A Server device and controlling method thereof Collects on-device AI results from refrigerators, diagnoses performance, ships model-update data Google Patents ↗
  3. KR20250032790A Refrigerator and controlling method thereof Food detection and tracking, in/out decisions, feature and image database Google Patents ↗
  4. KR20240109129A Method and apparatus for providing promotions using reinforcement learning Purchase-intent-driven product bundles and discount rates Google Patents ↗
  5. KR20230032843A Electronic device and controlling method of electronic device SR-GCL session-graph augmentation and recommender training Google Patents ↗
  6. KR20220052779A Electronic device and controlling method of electronic device Vision-AI image outpainting with segmentation Google Patents ↗

Publications

Newest first · 3 published · 2 under review

Google Scholar profile ↗

  1. Human Face 3D Reconstruction with Handheld Single 2D Camera on Mobile Devices IEEE ICCE 2016 DOI ↗
  2. Multi-Device Framework with Spatial Clustering ACM Middleware 2019 Industry Track DOI ↗
  3. SR-GCL: Session-Based Recommendation with Global Context Enhanced Augmentation in Contrastive Learning DLG-AAAI 2022 arXiv ↗
  4. From Selection to Switching in RLVR ICLR 2027 submission · under review
  5. KDD 2027 submission Under review · title withheld

Talks & Media

Newest first · 3 media appearances · 4 talks delivered

  1. AI 학습의 다음 단계, ‘스스로 만드는 데이터’ MIT Technology Review Korea · feature on self-evolving data Article ↗
  2. Self-Evolving Data in Physical AI NdotLight · Physical AI Night @ Humanoids Summit Seoul · delivered Event ↗
  3. AI Home Appliances = Samsung Samsung Newsroom · AI Vision Inside feature YouTube ↗
  4. On-Device AI That Will Reinvent Home Appliances AI Summit Seoul & Expo 2025 · AI Tech Lead Event ↗
  5. Inside Samsung: AI Vision Inside Bloomberg Media Studios · developer interview YouTube ↗
  6. Orchestrating On-Device AI and AI Agents Cloud & AI Summit 2026 Program ↗
  7. Self-Evolving Data: How AI Learns to Create Its Own Data AI Summit Seoul & Expo 2026 · delivered Program ↗

Personal Projects

Tools and learning sites I build and run outside work, designed, implemented and tested with several coding agents working in parallel. They run privately on my personal site, 33modeling.

  1. 2026.10–present App Storyboard A tool for designing the screens, flows and states of Android and iPhone apps. Layers are edited on a Figma-like canvas with a frame for each screen, and the design is exported as implementation instructions for AI and as starter code.
  2. 2026.09–present Generative tools from storyboard to picture book and video Storyboards plan each scene's text and image or video prompts. Character sheets keep characters consistent across illustrations, and scenes become narrated video clips joined into one film. Character settings defined once carry through from the storyboard to every image and video prompt.
  3. 2026.08–present 300+ learning sites and apps History, computer science, AI, school subjects, languages, science and early learning, with textbook-level lessons, worked examples, exercises and review.

Contact

33modeling@gmail.com ↗ minsoo3.kim@samsung.com ↗ GitHub ↗ Google Scholar ↗ LinkedIn ↗