Skip to content

Latest commit

 

History

History
138 lines (109 loc) · 7.4 KB

README.md

File metadata and controls

138 lines (109 loc) · 7.4 KB
Raw

Yongseop Kim

(updated: 2024-07-30)

Current Position

  • Employed at Samsung Electronics since 2014, contributing to advanced projects in AI and machine learning technologies.

Recent Work: Advanced Bitcoin Studies

I am currently deeply interested in blockchain technology, particularly Bitcoin, and am studying it in-depth.

Career

Development of LLM Service

  • 2024-05-01 ~ Present
  • Developing an LLM service for in-house use
  • Technologies used: Python, Litestar/FastAPI, PostgreSQL, LlamaIndex, LangChain, LangGraph

Development of Runtime for NN Models

  • 2023-12-01 ~ 2024-04-30
  • Developed a runtime (Neural Network Framework Accelerating Solution) for NN models
  • Implemented training features for x86_64 architecture
  • Repo: Samsung/ONE
  • Issues: Tasks, Activity
  • PRs: Tasks, Reviews

SW Challenge Project: HARP (Heterogeneous AcceleRator Platform)

  • 2022-07-16 ~ 2023-11-30
  • Designed and developed a scheduler framework for model training in a heterogeneous GPU environment (V100/A100)
  • Developed a prototype aimed at integration into the company's GPU infrastructure services
  • Technologies used: Go, GORM, Echo, Elastic, Python, Horovod, TensorFlow, PyTorch, Docker, MariaDB, Swit Templates Engine
  • Repo: in-house project

Development of a Visual Studio Code Extension for the ONE Compiler Toolchain

  • 2022-03-02 ~ 2022-07-15
  • Developed a Visual Studio Code extension for the ONE compiler toolchain
  • Core development: backend API and other interfaces
  • Set up test infrastructure
  • Technologies used: JavaScript, TypeScript, Mocha, VS Code Extension API
  • Repo: Samsung/ONE-vscode
  • Issues: Tasks, Activity
  • PRs: Tasks, Reviews

Development of NPU Chip Compiler for OnDevice AI

  • 2021-03-02 ~ 2022-02-28
  • Developed an NPU chip compiler for OnDevice AI
  • Defined IR and supported DSP (Digital Signal Processor)
  • Supported AI models for TV: Object Detection, Style GAN, etc.
  • Technologies used: C/C++, Python/Shell, GTest, TensorFlow, TensorFlow Lite
  • Repo: in-house project

Development of Runtime for NN Models

  • 2018-03-02 ~ 2021-02-26
  • Developed a runtime (Neural Network Framework Accelerating Solution) for NN models
  • Implemented a memory allocator for CPU/GPU and managed memory peak
  • Set up test infrastructure and supported features like subgraph/sequential, FP16, etc.
  • Supported commercialization through demos (Object Detection Model, GS7/Android App, E2EASR Model, GS20+/Android App)
  • Supported Odroid XU4, MuseM/Nike-M(TV) Tizen 5.5
  • Technologies used: C/C++, Python, Caffe, TensorFlow, TensorFlow Lite, Arm Compute Library, OpenCL
  • Repo: Samsung/ONE
  • Issues: Tasks, Activity
  • PRs: Tasks, Reviews

Development of DotNet Runtime

  • 2016-07-01 ~ 2018-02-28
  • Developed CoreCLR (Core Common Language Runtime) of .NET Core
  • Developed RyuJIT Compiler Codegen for ARM (Raspberry Pi3, TM1 (Tizen 3.0), TV Jazz-M/Kant-M (Tizen 3.0/4.0))
  • Developed LegacyJIT Compiler Codegen for x86/Linux
  • Specifically, supported PreCompile: Pre-compiling DLLs into Native Image (.dll) for better performance than JIT compile
  • Technologies used: C/C++/C#
  • Repo: dotnet/coreclr, dotnet/corefx, dotnet/runtime
  • Issues: runtime, runtime
  • PRs: coreclr, corefx

Development for Tizen OS

  • 2014-03-02 ~ 2016-06-30
  • Development for Tizen OS
    • Tizen Native C API
    • WRT (Web Runtime)
    • C# API
    • Xamarin

Technical Skills Summary

Programming Languages

  • Python: Extensive experience in backend development (FastAPI), test infrastructure setup, and AI model support (TensorFlow, TensorFlow Lite).
  • C/C++: Development of compilers, runtime environments, and memory management solutions for NN models and AI applications.
  • JavaScript/TypeScript: Development of Visual Studio Code extensions, backend APIs, and testing frameworks (Mocha).
  • Golang: Designed and developed scheduler frameworks for heterogeneous GPU environments.
  • C#: Development of .NET Core runtime and compiler code generation (RyuJIT, LegacyJIT).

Technologies and Frameworks

  • Machine Learning/AI: Implementation and support of NN models, AI frameworks (TensorFlow, TensorFlow Lite, Horovod), and neural network framework accelerating solutions.
  • Database Systems: PostgreSQL, MariaDB, and Elastic.
  • Containerization and Orchestration: Docker.
  • Web Development: FastAPI, Echo, GORM, and VS Code Extension API.
  • Test Frameworks: Mocha, GTest.

and

  • Blockchain: In-depth study of blockchain technology, particularly Bitcoin.

Development Tools

  • Build Systems and Compilers: Development of compiler toolchains and runtime environments for various architectures (ARM, x86).
  • Version Control Systems: Git, with extensive experience in managing repositories on GitHub.
  • Project Management and Version Control: Git, Jira, GitHub.
  • Containerization and Artifact Management: Docker and Nexus.
  • Continuous Integration/Continuous Deployment (CI/CD): JenkinsCI, GitHub Actions.
  • Integrated Development Environments (IDEs): Visual Studio Code.

Hardware Platforms

  • Heterogeneous GPU Environments: V100, A100.
  • Embedded Systems: Odroid XU4, various TV models with Tizen OS.
  • ARM Architecture: Raspberry Pi3, TM1 (Tizen 3.0), various TV models with Tizen 3.0/4.0.

Project Management and Collaboration

  • Open Source Contributions: Active participation in major open-source projects (Samsung/ONE, dotnet/coreclr, dotnet/corefx, dotnet/runtime).
  • In-House Development: Lead and contributed to various in-house projects focused on AI, machine learning, and runtime solutions.

Specific Domain Expertise

  • AI Model Support: Object detection, Style GAN, and other AI models for on-device applications.
  • Compiler Development: NPU chip compiler for on-device AI, .NET Core runtime and compiler development.
  • Memory Management: Implementation of memory allocators and peak management for CPU/GPU.