LUCID

Emergency medical AI for low- and middle-income countries.

A clearer
next step after
a head injury.

LUCID is a phone-based AI prototype designed to help families recognize potential warning signs after a head injury.

It asks structured questions, remembers earlier conversations, and calls back to check for changes—without an app or internet on the caller’s phone.

Explore the conversation View presentation

Functional prototype. Not a live medical service.

RECORDED PROTOTYPE

Hear LUCID in action.

About 2 minutesWatch on YouTube
✳1st place — mission:BRAIN Hackathon 2026Built by a University of Utah student team

01 / WHY LUCID

The first conversation
should not be
the last check.

After a possible head injury, a family needs to know what to do next. Finding guidance — and keeping track of what changes — can be difficult.

We’re focused on the gap between an injury and reaching care: making the next step easier to understand through an interface people already know. A phone conversation can gather context, continue after an interruption, and make room for a follow-up.

How the idea began

02 / HOW IT WORKS

A call. A conversation.
A way to stay connected.

Designed around the phone people already have — and the reality that a connection may not last.

01

A normal phone call.

An ordinary telephone, including a landline. No app or internet connection on the caller’s phone.

02

Structured questions.

Conversational AI gathers context through guided questions, with explicit triage and escalation rules.

03

A conversation that continues.

LUCID remembers earlier answers across interrupted calls and calls back to check for changes.

04

A path toward help.

The prototype demonstrates escalation prompts and routing where configured. Local referral pathways and clinical oversight are needed before use.

Language is part of access. Multilingual interaction is part of the prototype. Language needs and performance must be established for each intended setting.

03 / LUCID IN CONVERSATION

One conversation.
Even across calls.

An edited walkthrough based on our recorded prototype demo: a caller asks for clarification, LUCID remembers an interrupted call, and a follow-up checks for changes.

LUCIDCall connected
Ready to read
  1. The first call · Ask naturally
  2. LUCID

    Hi, I’m LUCID. I’ll ask a few quick questions about the person who hit their head. Did they lose consciousness?

CONCERN IN THIS DEMO · 1

Gathering information

LUCID begins gathering information about the head injury.

  • Loss of consciousnessAwaiting an answer
  • One-sided weaknessNot checked yet
  • Repeated vomitingNot shown yet
Rule state, not a risk score
READ ALONG1 / 20
LUCID: Hi, I’m LUCID. I’ll ask a few quick questions about the person who hit their head. Did they lose consciousness?

Adapted from the recorded demo, with brief connecting dialogue added. No live call or medical assessment.

Read the demo text Three short scenes

The first call

Ask naturally

  1. LUCID

    Hi, I’m LUCID. I’ll ask a few quick questions about the person who hit their head. Did they lose consciousness?

  2. CALLER

    Yes, they did.

  3. LUCID

    Is one side of their body weak, numb, drooping, or not moving normally?

  4. CALLER

    How could I test that? I’m not really sure.

  5. LUCID

    Have them squeeze your hands with both hands.

  6. CALLER

    It seems like it’s pretty even pressure. Is there any other test I can do?

  7. LUCID

    Ask them to lift both arms straight out in front of them. Does either arm drift down or look different?

  8. CALLER

    They look the same.

  9. Call dropped · connection lost.

After a dropped call

Pick up again

  1. The caller calls back. Reconnecting…
  2. LUCID

    The call dropped before we finished. You said they lost consciousness, with no weakness on one side. Is that still correct? Has anything changed?

  3. CALLER

    That’s still right. Nothing has changed.

  4. LUCID

    I’ll call back to check again.

  5. CALLER

    Thank you. Goodbye.

  6. Call ended. A check-in is scheduled.

Later · time compressed

Check back

  1. Later · incoming call from LUCID.
  2. LUCID

    Hi, it’s LUCID calling back about the head injury. Have they vomited more than once since the injury?

  3. CALLER

    They did vomit one more time.

  4. LUCID

    They need to get to a hospital now.

  5. CALLER

    Understood.

This adapted conversation demonstrates prototype behavior, not a clinical protocol. General information: CDC head-injury danger signs ↗.

04 / OUR STORY

How the idea is taking shape.

A question from the hackathon led to a working prototype. We’re now exploring what it would take to adapt that approach with local partners.

  1. 01 / THE STARTING POINT

    Knowing when to seek help.

    At the 2026 mission:BRAIN hackathon, a fictional case in rural Gilgit-Baltistan, Pakistan, described a family facing uncertainty after a head injury and a difficult journey to care. We focused on one gap: how could they recognize the need for help before reaching a specialist?

  2. 02 / WHAT WE BUILT

    A conversation that continues.

    Our response was LUCID: a functional conversational AI prototype using an ordinary phone call for guided questions, remembered context, and follow-up. The project won first place at the hackathon.

  3. 03 / WHAT WE’RE EXPLORING

    A locally shaped next step.

    With Glial Initiative, we’re using the prototype as the starting point for a Nigeria-specific proposal for Wema Hackaholics’ Social Impact track. Local research will shape what needs to change, from language and telephone access to referral pathways and the operating model.

Local knowledge comes first.

The immediate work is an implementation and sustainability proposal, supported by an adapted demonstration.

LUCID develops the software. Glial contributes local research, community engagement, and stakeholder introductions. Both teams shape the operating and funding model.

  1. Who would use it, and in which languages?
  2. How would calls connect to local care and referral pathways?
  3. What would evaluation, operation, and funding require?

Clinical review and evaluation planning are needed before deciding whether to pursue a supervised pilot.

05 / THE TEAM

A small team.
A practical question.

LUCID was created by three University of Utah students at the 2026 mission:BRAIN hackathon.

WS

Walter Shewmake

AI architecture & engineering

DR

David Ross

Clinical logic & decision systems

DC

Dennis Courtright

Business strategy & concept validation

A FEW ANSWERS

Where things stand.

Is LUCID available for medical use?

No. It is a functional proof of concept, not a live or clinically validated medical service. This website explains the project; it does not provide medical assessment.

Does it require a smartphone or internet?

The intended caller interface is a normal phone call, including from a landline. The caller does not need an app or mobile internet. The backend still requires computing and telecommunications infrastructure.

Has LUCID been used with patients?

No clinical field deployment or patient-facing service is established. The Pakistan origin is a fictional hackathon case. The proposed Nigeria work concerns local research, planning, and prototype adaptation; a community or pilot site has not been selected.

Which languages does it support?

Multilingual interaction is part of the prototype. The languages needed for a Nigerian setting and their performance will be assessed with local input.

How can clinicians or community organizations contribute?

We welcome conversations about local needs, language, referral pathways, and responsible evaluation. You can contact Walter using the link below. No funded pilot or formal partner program is established.

CONTINUE THE CONVERSATION

Help shape the next step.

We welcome conversations with clinicians, community organizations,
and potential implementation partners.

Contact View the presentation