[go: up one dir, main page]

Computational Linguist
Language, Cognition & LLM Alignment

Kai Park

Quantifying the nuance in how speakers mean more than they say, so language models can learn to follow.

About

Kai Park

(Hyeyeon Park)

Computational linguist working at the intersection of empirical linguistics and language model evaluation. I study the cognitive mechanisms behind how humans process and represent language, using corpus-based quantitative analysis and controlled experiments to test how linguistic structure interacts with cognitive constraints.

My current work applies these findings to language model evaluation and socio-affective alignment. I turn pragmatic, emotional and cultural nuance into measurable signals that show where model behavior aligns with human language and where it diverges. I own projects end to end, from experimental design to publication, and sharpen them through supervisors, colloquia and peer review.

Working across Korean, English and German gives me a fast start for pilot studies. The longer aim is scalable alignment methods that hold across linguistic structures, not just within one language.

Interests
AI alignment, natural language processing
Ongoing
Linguistic alignment, approached from several angles: where model language departs from how people actually use it.
Programming
Python and R for research (data pipelines, statistics) · Java and Prolog from coursework
Prototyping
Demos and prototypes with coding agents (Claude Code, Antigravity). This site is one of them ;)
Beta testing
I try new AI products early and send structured feedback.
Based
Germany · authorized to work (post-study permit), no visa sponsorship needed · open to relocation
Availability
From January 2027, for full-time research roles in industry or academia, and for PhD positions

Recent work (sole-authored)

All research →

I enjoy thinking these questions through with others. If any of this overlaps with your work, I'd be glad to talk.

Recent

News

  • Nov – Dec 2026Attending the Frontier Research Club series "Beyond the Bench: Engineering Autonomous AI Research Loops" in Munich (9 Nov) and Zurich (3 Dec), hosted by Digital Schooling and Google Cloud.
  • Oct 2026The Grue Divide presented as a poster at PANDORA @ EMNLP 2026.
  • Sep 2026Submitted MA thesis (11 Sep) on morphological derivation and the Korean blue-green spectrum, Göttingen.
  • Apr 2026Selected for the Google Cloud TPU Research Cloud program.

All news →