I'm an AI researcher focused on turning research ideas into working systems quickly.

Research ideas should survive contact with code.

My current interests sit around multimodal models, speech, LLM evaluation, paper tooling, and the practical gap between "this might work" and "here is a working prototype." I like research that is ambitious enough to matter and concrete enough to run.

• Building taste as an AI researcher: reading, surveying, reproducing, and asking better questions.
• Shipping faster: turning rough research ideas into small tools, demos, and experiments.
• Working around speech, multimodal LLMs, document understanding, and research automation.

• SilVar
  EMNLP 2025 work on reasoning speech instruction with large visual-language models for object localization and text generation. This is the biggest thing I have worked on so far.

• Speech_Tokenizer_Lib
  A reference collection of speech tokenizers that matter for speech model development.

• AI-Conferences-Journals-Papers and Cannot-Miss-AI-Conferences-Journals
  Tools and notes for staying oriented in the AI research literature.

• crawl-ref-papers-gg-scholar and paper-cli
  Research workflow tools for crawling, indexing, downloading, and searching papers.

• from_scratch
  My place for rebuilding AI ideas from scratch, dependency-light, to understand what is really going on.

• RAG-basic-system
  A small RAG system for practicing the path from concept to implementation.

I use GitHub as a research workshop: some repos are polished, some are experiments, and some are receipts from learning in public. The direction is becoming sharper over time:

1. Find an important research idea.
2. Understand it deeply enough to reproduce or stress-test it.
3. Build the smallest useful version.
4. Keep what compounds; archive what was only a stepping stone.

Tiny prototypes count. So do careful surveys. So does deleting messy stuff when it stops helping.

• Youtube-Download - a practical downloader tool.
• useful-tools - local-first tools I use or study.
• python-coding-stuff - Python notes and small utilities.
• LaTeX - my LaTeX workspace.

• GitHub: christian-hoang-04
• Email: christianhoang04@gmail.com

If a repo looks like a lab bench, that is probably because it is one.