
Peter James
Co-founder, ex-Google Product Lead and Engineer. Leads technical study design and system architecture. Engineer-researcher in real-time multimodal AI for clinical contexts.
Patients simply speak; our system generates the exact spoken explanation they need to hear, the visual they need to see, and a clinician-reviewable record. We study how this works in production - and publish what we learn.

Read our original Emperor’s Interface essay
Our Research Mission
Academic and laboratory teams rarely have access to large-scale, longitudinal, multimodal interaction data from real users. LinkaChart operates as a living laboratory: a deployed product used daily by consenting patients with complex chronic illness, so the research community can study real-time micro-turn steering, generative UI, and live agent control in the wild. The product surface we study, Capture, Show, Guide, is one continuous story across three steps: between, during, and after appointments. Our protocols and outputs are deliberately domain-agnostic, designed to be useful well beyond chronic care.
Current Research Focus
Extending mid-conversation tool calls (dynamic next-question selection, instant symptom logging) into full micro-turn steering with context injection and live redirection.
Orchestrated agents that produce the right spoken explanation and the right visual - charts, symptom timelines, source references - in real time, on top of an open-weights base model.
Randomized, clinician-blinded head-to-head study comparing the micro-turn Emperor’s Interface against a turn-based baseline, with ablations isolating acoustic features, sentiment signal, and policy specialization.
Tracking state integrity, symptom recall, and safety-floor flags across consented patients over a multi-week longitudinal window - not a single-session snapshot.
Outputs will be released publicly on completion.
Three steps, one continuous story
Capture
Generative visual + spoken response.
Show
Structured summary your care team can actually use.
Guide
Doctor's instructions into daily guidance.
Open Research Commitment
We view this work as filling a structural gap in the literature: results any researcher can understand, validate, and use - not just specialists. Outputs are released under permissive open licences on completion.
Phase 1
Acoustic feature extraction, prototype steering extensions, clinician rubric pilot with inter-rater reliability testing, and enrollment.
Phase 2
Comparative study results, Doctor Rubric with IRR calibration process, arXiv pre-print, and peer-review submission.
Research Team

Co-founder, ex-Google Product Lead and Engineer. Leads technical study design and system architecture. Engineer-researcher in real-time multimodal AI for clinical contexts.

Co-founder, Physician. Leads clinical protocol design, patient safety reviews, and floor-level oversight for high-acuity care.
Advisors & collaborators - we work alongside independent clinicians, academics, and ethics reviewers.
Get Involved
Open to LinkaChart users via opt-in consent. Data is encrypted, de-identified, and re-identification risk is addressed before release. Care is unaffected; withdraw any time.
Interested in participating in research? Share a few details and we’ll be in touch about studies you may qualify for.
Interested in partnering on research? Tell us about you and what you’d like to explore together.