GlucoCoach
Inspiration
Type 2 Diabetes is everywhere in medicine, yet students often learn it in fragments. One lecture explains insulin resistance. Another covers beta cell failure. Later, complications appear. Pharmacology feels disconnected.
We kept asking:
Why are we memorizing mechanisms in isolation when the disease itself is systemic and interconnected?
We wanted to build something that connects physiology to clinical decision-making in one continuous experience. First understand what is happening inside the body. Then apply that understanding to real patient scenarios.
That became GlucoCoach.
What It Does
GlucoCoach has two tightly connected interfaces.
1. The Physiology Explorer
This is the interactive learning phase.
Users manipulate:
- Carbohydrate intake
- Activity level
- Insulin sensitivity
- Inflammation
- Chronic exposure
They watch glucose and insulin curves shift in real time. A dynamic body map updates organ risk as physiology changes. Clicking an organ reveals why complications occur.
Everything is cause and effect.
An integrated AI tutor, GlucoCoach, guides learners through the mechanisms. It explains physiology conversationally and remains strictly scoped to Type 2 Diabetes.
Goal: deep mechanistic understanding.
2. The Clinical Decision Interface
After mastering mechanisms, users enter a patient case simulation.
Each case includes:
- Patient history
- Comorbidities
- Risk factors
- Lab values
- Current symptoms
Instead of memorizing drug lists, users drag and drop medication classes into the patient profile.
The system evaluates reasoning across:
- Glycemic control
- Cardiovascular benefit
- Renal impact
- Weight implications
- Side effect risk
- Cost and practicality
If a choice is inappropriate, the affected domain is visually flagged and explained.
Pharmacology becomes structured reasoning instead of recall.
Together, these interfaces connect physiology directly to treatment logic.
How We Built It
GlucoCoach is a modular web platform built on a modern React-based architecture.
Core Systems
Deterministic physiology engine Based on structured datasets to ensure stable, explainable outputs.
Rule-based clinical reasoning engine Built on structured treatment guidelines. Drug classes influence defined physiological and risk parameters.
Reusable components for:
- Interactive body maps
- Dynamic glucose and insulin graphs
- Organ-level complication visualization
- Case-based medication reasoning
- A scoped AI tutoring engine
All thresholds, ranges, and explanatory logic are stored in structured local data packs derived from credible sources.
The architecture is modular, allowing new cases, drug classes, or conditions to be added easily.
Challenges
Stability and Intuition
Early physiology models felt unpredictable. We rebuilt the engine so small input changes produce logical, explainable shifts.
Avoiding Oversimplification
Clinical decisions are nuanced. We designed rule systems that reflect guideline reasoning without overwhelming the learner.
Safe AI Boundaries
Designing GlucoCoach required strict scope control to prevent drift into unsafe medical advice.
Balancing realism with usability was the most difficult design challenge.
What We’re Proud Of
- Theory and practice are not separated.
- Users see why glucose rises before choosing how to treat it.
- Medication selection explains consequences across multiple domains instead of labeling answers right or wrong.
- GlucoCoach enhances understanding without replacing reasoning.
- The system feels cohesive: physiology flows naturally into pharmacology.
What We Learned
- Understanding improves when learners manipulate variables and observe immediate outcomes.
- Clinical reasoning can be structured and visualized.
- AI is most powerful when it is focused and bounded.
- Translating complex medical systems into intuitive digital experiences is harder than it appears.
What’s Next
We plan to:
- Expand the case library with varied patient profiles
- Add deeper pharmacologic mechanism modeling
- Introduce adaptive AI to identify misconceptions
- Integrate anonymized real-world datasets
- Expand beyond Type 2 Diabetes into metabolic syndrome and cardiovascular disease
Long-Term Vision
GlucoCoach will become a unified platform where students, trainees, and clinicians learn pathophysiology and clinical reasoning in one continuous, connected system.
Log in or sign up for Devpost to join the conversation.