📌 Overview

AI Systems Engineer specializing in end-to-end AI system development—from lightweight model design to scalable infrastructure deployment. Experienced in deploying ultra-lightweight thermal object detection on MCUs and designing secure server architectures for air-gapped environments.

Highlights:

• Published as first author in IEEE Access (SCIE) and presented two papers at IEEE ICCE 2025
• Filed 2 domestic patents and 1 international PCT patent
• Experience across the full stack: AI models, embedded systems, backend services, and cloud infrastructure

ORCID: 0009-0004-0998-6643

---

📌 Professional Experience

A.I.Matics — AI Systems Developer (Sep 2025 ~ Present)

Tech Stack: C/C++, Python, TensorRT, TFLite, Docker, Embedded Linux, OpenCV

I optimize AI models for edge devices and design system architectures for real-world deployment.

Key Projects

Mobile-Conn Gateway Application — Deployed on Roadscope R11

• Built a C++ gateway exposing internal system state via HTTP(S) REST API
• Implemented ZeroMQ IPC, cpp-httplib REST server, and HTTPS certificate security
• Delivered APIs for camera/ADAS/sensor configuration, firmware management, and device provisioning
• Successfully deployed in production

Zero-Touch Server Architecture for Restricted Networks — [Patent Pending: 1st Inventor]

• Designed an automated deployment system: USB insert → full service ready in ~17 minutes
• Key technologies:
  • cloud-init + autoinstall for zero-touch OS and GPU setup
  • LXC + Docker nested virtualization for complete service isolation
  • L2 network integration with direct DHCP/ARP and mDNS support
  • WireGuard/Firewall auto-configuration and TPM-based integrity checks
• Built offline package pipelines for air-gapped environments
• Engineered for manufacturing deployment (Dreamtech environment)

---

한국인물검증(Korea Personnel Verification Company)

Backend Developer (Dec 2021 ~ Dec 2022)

Tech Stack: Spring Boot, Flask, AWS, Swagger, Jenkins

Project: KKAYO — Personal Information Verification System

• Built diploma verification engine and Spring Boot API server
• Managed AWS cloud infrastructure
• Results: 3x faster response times, 30% reduction in server costs

---

📌 Research

FLARE: Real-Time Thermal Object Detection (2024)

Developed an ultra-lightweight detection model for low-resolution thermal cameras.

• 118x smaller memory footprint and 12x faster than YOLOv8n
• Runs in real-time on STM32 MCU
• Custom augmentation techniques achieving 95% mAP

Tech: Python, TFLite, OpenCV, YOLO, MobileNetV2-SSD, STM32, X-CUBE-AI

---

ObjectBlend: Data Augmentation for Imbalanced Datasets (2024)

Created a novel augmentation technique for industrial defect detection.

• Improved YOLOv3-tiny mAP50-95: 0.20 → 0.91
• Outperforms CutMix and similar methods
• Significantly boosts minority class (defect) detection

Tech: Python, OpenCV, PyTorch, YOLO, NVIDIA AGX Orin, TensorRT

---

📌 Publications

Journal

• IEEE Access (SCIE) — Real-Time Object Detection Using Low-Resolution Thermal Camera for Smart Ventilation Systems
  DOI: 10.1109/ACCESS.2025.3566635

• IEEE Access (SCIE) — ObjectBlend: Data Augmentation Technique for Vision Inspection Systems (Under Review)

Conference

• IEEE ICCE 2025 — Real-Time Object Detection Using Low-Resolution Thermal Camera
  DOI: 10.1109/icce63647.2025.10930159

• IEEE ICCE 2025 — ObjectBlend: Data Augmentation Technique for Vision Inspection Systems
  DOI: 10.1109/icce63647.2025.10929866

---

📌 Patents

Korea

• Low-Resolution Thermal Object Detection and Real-Time Inference — 10-2024-0127351
• Thermal Imaging Data Augmentation Algorithm — 10-2024-0127352

International (PCT)

• Autonomous Ventilation System Using Predictive Model — PCT/KR2025/014712

---

📌 Skills

Category	Technologies
AI/ML	TensorFlow, PyTorch, YOLO, MobileNetV2-SSD, TFLite, TensorRT
Embedded	STM32 (CubeMX, X-CUBE-AI), Raspberry Pi, NVIDIA Jetson
Backend & Cloud	Spring Boot, Flask, FastAPI, AWS, Docker

---

🔗 GitHub: github.com/MartenLabs
🔗 Blog: MartenLabs