Looks like you have JavaScript disabled. For the full Twine experience, you will need to re-enable it.

I’m an AI/ML Engineer specialising in Computer Vision, Multimodal Learning, and Workflow Automation. At SSG.COM, I fine-tuned and deployed vision–language models (VLMs) for image–text retrieval and built SSG Lens, a real-time multimodal search service used in production. I also created a synthetic data generation pipeline (500K+ pairs) using LLM prompt engineering and generative models, reducing manual labeling by over 200× and accelerating downstream model training. I focused on optimizing inference performance (TF-TensorRT integration, latency 0.66s → 0.19s) and reducing vector dimensions by 75% for efficient large-scale retrieval. Before that, I worked at kakaoVX, developing gesture recognition and biometric authentication systems for sports and virtual experience products, and at Xinapse, where I built NLP/NLU chatbots and a bankruptcy prediction system for financial clients. I enjoy building practical AI systems that connect research to real-world products and exploring how AI and automation can make creative work more efficient.…I’m an AI/ML Engineer specialising in Computer Vision, Multimodal Learning, and Workflow Automation. At SSG.COM, I fine-tuned and deployed vision–language models (VLMs) for image–text retrieval and built SSG Lens, a real-time multimodal search service used in production. I also created a synthetic data generation pipeline (500K+ pairs) using LLM prompt engineering and generative models, reducing manual labeling by over 200× and accelerating downstream model training. I focused on optimizing inference performance (TF-TensorRT integration, latency 0.66s → 0.19s) and reducing vector dimensions by 75% for efficient large-scale retrieval. Before that, I worked at kakaoVX, developing gesture recognition and biometric authentication systems for sports and virtual experience products, and at Xinapse, where I built NLP/NLU chatbots and a bankruptcy prediction system for financial clients. I enjoy building practical AI systems that connect research to real-world products and exploring how AI and automation can make creative work more efficient.

Junghyun Park





I’m an AI/ML Engineer specialising in Computer Vision, Multimodal Learning, and Workflow Automation. At SSG.COM, I fine-tuned and deployed vision–language models (VLMs) for image–text retrieval and built SSG Lens, a real-time multimodal search service used in production. I also created a synthetic data generation pipeline (500K+ pairs) using LLM prompt engineering and generative models, reducing manual labeling by over 200× and accelerating downstream model training. I focused on optimizing inference performance (TF-TensorRT integration, latency 0.66s → 0.19s) and reducing vector dimensions by 75% for efficient large-scale retrieval. Before that, I worked at kakaoVX, developing gesture recognition and biometric authentication systems for sports and virtual experience products, and at Xinapse, where I built NLP/NLU chatbots and a bankruptcy prediction system for financial clients. I enjoy building practical AI systems that connect research to real-world products and exploring how AI and automation can make creative work more efficient.…I’m an AI/ML Engineer specialising in Computer Vision, Multimodal Learning, and Workflow Automation. At SSG.COM, I fine-tuned and deployed vision–language models (VLMs) for image–text retrieval and built SSG Lens, a real-time multimodal search service used in production. I also created a synthetic data generation pipeline (500K+ pairs) using LLM prompt engineering and generative models, reducing manual labeling by over 200× and accelerating downstream model training. I focused on optimizing inference performance (TF-TensorRT integration, latency 0.66s → 0.19s) and reducing vector dimensions by 75% for efficient large-scale retrieval. Before that, I worked at kakaoVX, developing gesture recognition and biometric authentication systems for sports and virtual experience products, and at Xinapse, where I built NLP/NLU chatbots and a bankruptcy prediction system for financial clients. I enjoy building practical AI systems that connect research to real-world products and exploring how AI and automation can make creative work more efficient.

Available to hire

I’m an AI/ML Engineer specialising in Computer Vision, Multimodal Learning, and Workflow Automation.
At SSG.COM, I fine-tuned and deployed vision–language models (VLMs) for image–text retrieval and built SSG Lens, a real-time multimodal search service used in production.
I also created a synthetic data generation pipeline (500K+ pairs) using LLM prompt engineering and generative models, reducing manual labeling by over 200× and accelerating downstream model training.
I focused on optimizing inference performance (TF-TensorRT integration, latency 0.66s → 0.19s) and reducing vector dimensions by 75% for efficient large-scale retrieval.
Before that, I worked at kakaoVX, developing gesture recognition and biometric authentication systems for sports and virtual experience products, and at Xinapse, where I built NLP/NLU chatbots and a bankruptcy prediction system for financial clients.
I enjoy building practical AI systems that connect research to real-world products and exploring how AI and automation can make creative work more efficient.

Skills

Experience Level

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Expert

Language

Korean

Fluent

English

Advanced

Work Experience

Machine Learning Engineer at SSG.COM (E-commerce)

May 1, 2025 - May 1, 2025

Led end-to-end ML development for image-text retrieval and multimodal search. Fine-tuned and deployed vision-language models (VLMs) enabling content discovery and personalized recommendations. Improved semantic alignment between product images and user queries, increasing retrieval accuracy in production. Built a synthetic dataset generation pipeline (500K+ pairs) using LLM prompting and generative models, reducing manual labeling effort by 200x and accelerating downstream model training. Developed SSG Lens, a large-scale real-time image search service with multi-task multimodal embeddings across 13 fashion attributes (color, texture, pattern). Reduced vector dimensionality by 75%, improving retrieval efficiency across millions of products. Optimized inference with TF-TensorRT, cutting response latency from 0.66s to 0.19s. Built Gradio-based rapid prototypes for image-captioning pipelines and stakeholder feedback. Led cross-functional efforts in developing and deploying multimodal sear

Machine Learning Engineer at KAO VX (Sports & Virtual Experience)

May 1, 2021 - May 1, 2021

Developed a gesture recognition system for home fitness apps, achieving real-time 30 fps inference and enabling hands-free device navigation. Implemented low-latency biometric authentication (palm print recognition) optimized for mobile deployment. Prototyped a GAN-based weather simulation system for virtual golf to enhance immersion. Delivered an automated sentiment analysis pipeline for customer reviews, improving service planning efficiency.

Machine Learning Engineer at Inapse (Financial AI Startup)

November 1, 2018 - November 1, 2018

Developed an NLP/NLU framework powering university and election chatbots, scaling to thousands of users. Built a bankruptcy prediction system using statistical modeling, regression, and tree-based methods for early warning and credit risk management, supporting consultations with the Export-Import Bank on distressed companies.