Projects

A collection of my work spanning iOS development, web applications, open-source tools, and research projects.

Filter by:AVFoundationAstroAudio EngineeringDeep LearningDeepSpeedFigmaFrontendGPU ComputingHPCLLMsLoggingMedia Visualization

All Projects

DubDubSings icon

DubDubSings

Make Siri sing your favorite song!

Initially introduced as SiriSings for my WWDC23 Swift Student Challenge Submission, DubDubSings serves as an interactive platform for aspiring songwriters. It utilizes Apple's AVSpeechSynthesizer to convert text into spoken words and bring musical ideas to life with expressive audio controls.

  • WWDC23 Swift Student Challenge Submission
  • Featured on WWDC Scholars
  • Published on App Store
SwiftSwiftUIiPadOSText-to-SpeechAudio EngineeringAVFoundation+1
Archived
MusicBar icon

MusicBar

Easy, Lightweight Status Bar Media Visualizer

inedible.dev

A macOS lightweight status bar media visualizer that integrates directly into the macOS status bar. MusicBar provides instant visual feedback on the current media without disrupting your workflow, featuring real-time audio visualization and seamless integration with Apple Music and Spotify.

  • Built with SwiftUI for native macOS integration
  • Real-time audio visualization
  • Supports Apple Music and Spotify
SwiftmacOSSwiftUIFrontendAstroMedia Visualization+1
mixr icon

mixr

Lightweight Virtual Mixer

My WWDC22 Swift Student Challenge Submission - a lightweight virtual mixer designed for everyone from beginners to mix engineers. It allows for referencing music on specific Apple devices and provides individual control over each track or instrument in a song.

  • WWDC22 Swift Student Challenge Submission
  • Featured in App Store Story
  • Individual track/instrument control
SwiftiPadOSXcodeAudio EngineeringFigmaWWDC22+1
Completed

Scaling on LANTA

Driving Multi-GPU Efficiency for LLMs

NSTDA Supercomputer Center (ThaiSC)

Optimizing Large Language Model training through scaling techniques. This project explores weak and strong scaling in LLM training using DeepSpeed, Flash Attention 2, and LoRA. We assessed multi-node efficiency, memory usage, and runtime optimization using NCCL configurations, developing custom tools for performance monitoring, fine-tuning automation, and visualization.

  • Showcased at AI Thailand Forum 2024
  • Multi-GPU training optimization
  • Custom performance monitoring tools
HPCLLMsScalabilityDeepSpeedNCCLDeep Learning+3
Completed

mtnlog

Simple Multinode Performance Logger for Python

NSTDA Supercomputer Center (ThaiSC)

A simple multinode performance logger for Python, designed to be used similarly to Python's built-in logging module but with a focus on performance logging. It provides a simple API for logging performance data including start and end times, and allows for easy integration with other logging systems.

  • Published on PyPI for easy installation
  • Simple API similar to Python's logging module
  • Designed for HPC environments with Slurm
PythonSlurmHPCLoggingPerformance Monitoring
5
Total Projects
28
Technologies Used