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

#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.

#TensorFlow Lite #Python #Interpreter

• Works on TensorFlow Lite Python Interpreter.
  Based on: https://www.tensorflow.org/lite/examples, converted Android Java(Kotlin) code to Python
  
  • TFLitePoseEstimation
  • TFLiteDetection
  • TFLiteClassification
  • TFLiteSegmentation

#Mobile #PM #flask #svm #dark mode

#HighCharts.js #Go.js #flask #csv #Web browser