AI products and research

AI products for real work, grounded in research.

transkinesis builds domain AI products — a CRM for Korean medicine (한의원) clinics and a local-first research workspace — backed by machine-learning research in mobility and language.

Approach

How the products are built.

  • 01

    Domain-aware, not generic

    Each product is shaped around a real practice — a clinic's patient flow, a researcher's reading loop — instead of a one-size assistant.

  • 02

    Human-in-the-loop by design

    AI prepares the next move: summaries, drafts, next actions. People keep control of records, decisions, and approvals.

  • 03

    Local-first and private

    Sensitive data — patient records, unpublished research — can stay on your own machine or infrastructure.

  • 04

    Research-backed

    The same person building the products publishes the methods behind them — established work in mobility ML, with emerging work on language models.

Research

Spatio-temporal machine learning for how cities move.

The products sit close to active research: forecasting, origin-destination modeling, and learning under structure for mobility systems — with emerging work on language models for research and decision workflows.

View research themes
  • 01

    Spatio-temporal forecasting

    Forecasting models that treat time, geography, and mobility demand as connected structure.

  • 02

    Origin-destination modeling

    Methods for representing and predicting multi-modal movement patterns across urban systems.

  • 03

    Learning under structure

    Lipschitz-aware learning, graph structure, and transport constraints for stable modeling.

  • 04

    Language models for research

    Emerging work applying and evaluating language models for research workflows and structured decision support — the direction behind Local Research.

Have a workflow that doesn't fit inside off-the-shelf software?

transkinesis is open to product pilots, research collaborations, and applied AI work — in Korean and English.

Start a conversation