I received the M.Sc. degree in 2021 from Yonsei University, Seoul, Korea. Currently interested in designing MLOps architecture with IaC especially AWS CDK.
My CV

#Multimodal Embedding#Fused Search

• Multimodal Search for Devsisters
  • Developed a Multimodal Embedding Search for over 1 million in-house art assets.
  • Supported full-text and vector fusion search to maximize search relevance for planners and artists.
  • Implemented clustering functionality to intuitively group and display similar search results.
  • Currently optimizing vector search algorithms to reduce latency and improve overall retrieval speed.

#AI Agent#Skills#Tools#SubAgents

• AI Agent for Art Generation
  • Building an AI Agent designed to generate art assets strictly adhering to Devsisters' Cookie Run IP art style.
  • Generating various in-game assets, including backgrounds, cookie characters, and Cookie Run: Kingdom decorations.
  • Developing a conversational chat interface to provide a seamless and interactive UI/UX for internal users.

#Airflow#Triton#FastAPI#vLLM#MLOps#AI#VLM#SLM

• Synapsego: Automated Visual Synopsis Creation
  • Audio transcription model serving with Ray Serve
  • VLM multimodal inference serving and LLM/SLM serving with vllm + FastAPI
  • Vision models serving with Triton BLS
  • Pipelining with Airflow DAGs

#AWS#CDK#Serverless#IOS#AI#Agent

• shot up: AI assistant for screenshots
  • App Store Link
  • AWS architecture.
  • Analyzing with AI Agent
  • Semantic search.
  • AWS CDK(IaC) for deploying AWS resources.

#AWS#CDK#Serverless#Event-driven#IOS#AR

• aug: spacial social
  • App Store Link
  • AWS serverless event-driven architecture.
  • Video streaming & image content distribution.
  • AWS CDK(IaC) for deploying AWS resources.

#AWS#CDK#MLOps#AI#CCTV

• MLOps for AI Surveillance Camera

  • Architecture Diagram
  • Dataset pipeline using AWS Fathom(co-developing service with SKT)
  • Design & implement train, inference, conversion, deploy pipeline
  • Video streaming & image content delievery
  • Semantic image search
  • AWS CDK(IaC) for deploying AWS resources.

• Radio tower anomaly detection with drone images

  • Design serverless architecture for batch inference pipeline.
  • AWS CDK(IaC) for deploying AWS resources.

#linux#libuv#C++#C##Unity

• LibuvGameServer
  • On Ubuntu 18.04, using libuv for TCP connection.
  • Based on libuv's event loop, Network IO is single threaded.
  • [Bug] Segmentation Fault after continuous Disconnection and Connection.
• LibuvGameServerClients
  • Dummy Client is based on C# .NET framework
  • Client is based on Unity
  • Demo avaiable here

#mediapipe#blazepose#Unity#3D#C##Android

• BlazePoseWithUnity
  • Only has Demo and brief description of project due to confidential rights

#Face Recognition#Android#Firebase ML Kit#TensorFlow 2.0#TensorFlow Lite#Java#Python

• JHFace

  • Face recognition training & testing framework implemented with TensorFlow 2, Keras
  • Supported backbones: MobileNet, MobileNetV2, InceptionResNetV2, InceptionV3, ResNet50, ResNet50V2, ResNet101V2, MnasNetA1, MnasNetB1, MnaseNetSmall, NASNetLarge, NASNetMobile, Xception, MobileNetV3Large, MobileNetV3Small, EfficientNetLite0 ~ Lite6, EfficientNetB0 ~ B7
  • Supported losses: ArcFace, CosFace

• Android-FaceRecognition

  • Is runtime face identification on Android device. I used IJB-C dataset for testing labels.

• FaceBird

  • Is game applicaiton which utilize ML Kit for controlling the bird with rotating Face

#Gaze Tracking#Android#Firebase ML Kit#TensorFlow Lite#PerCom 2021#GAZEL#Java#Python

• GAZEL

  • Is a Personalized Runtime Mobile Gaze Tracker.
  • This work is official implementation of GAZEL framework which is published in PerCom 2021(GAZEL: Runtime Gaze Tracking for Smartphones) .

• GazeBird

  • Is game application which utilize GAZEL for controlling the bird.

• MLKitGazeDataCollectingButton

  • Is gaze data collecting application for making gaze estimation model required for GAZEL.