About
Research Interests
Current:
- Anti-MLsys: Detect and understand LLM-generated web content.
- Efficient LLM agents for the Web/ programming through domain knowledge.
Wider areas:
- Understanding the Web’s structures and contents.
- Efficient and high-performance computing with CPU cache optimization.
- Robust and user-friendly distributed computing like for websites and databases.
- Programming language infrastructure to enable robust static analysis.
Previous: Federated learning, Internet routing, content provenance (C2PA).
Contact
- Email: See the address in my CV. I get notifications and reply.
- I block whoever sends spam immediately.
- Ping me if I do not reply in three days.
- GitHub: Ping me for code-related discussions. I get emails.
For Undergrads/MS
I am looking for you if you want to 1) participate in PhD-level research (≠ want to do a PhD) and b) are suitable for working on something we are both interested in. I teach, so don’t worry about lacking specific skills.
You are welcome to email me succinctly to discuss your interests and suitability.
Necessary qualities:
- Ambition: You always want to achieve something greater and self improve.
- Dedication: If you do something, you do it well.
- Availability: You can spend lots of high-quality time on research.
Necessary skills (learnable):
- Learning: You learn new things quickly.
- Language: You can input (read/listen) and output (write/speak) technical and proper English (≠ technical papers), including instructions and docs of libraries.
- Computer: You are fluent in typing, a text editor, a scripting language, a systems language, and are generally familiar with basic OS terms.
Very helpful skills (non-exhaustive):
- You can implement complex systems.
- You can analyze & visualize messy domain-specific data.
- You can do literature review.
- You can read technical papers.
Name
The legal name is “Sichang He” (何思畅, or 何思暢). Steven is the preferred English name. The name used in publications is Sichang “Steven” He.
Name pronunciation
Google Translate pronounces it correctly (Click Listen). Serializing it into English is impossible; the best you can get is H-err? Sii↑ chung↓.