University healthcare technology prototype

GlucoFinity

Discover the possibilities within your glucose data.

GlucoFinity combines glucose readings with meals, sleep, activity, and medication data to help users recognize meaningful patterns in their health.

Explore the Demo How It Works

Educational prototype. Not intended for diagnosis, treatment, or medication decisions.

Today's overview

Fictional demonstration data

Demo sensor connected

Current glucose

118mg/dL

Within demonstration target range

Time in range

84%

70-180 mg/dL

Daily average

126

mg/dL

24-hour glucose trend

Shaded area shows the demo target range

Recent meal

Salmon rice bowl

12:35 PM

Sleep

7 h 18 min

Good continuity

Exercise

24 min walk

1:18 PM

Medication

Event logged

8:05 AM

AI-assisted insight

Prototype

Your glucose response after similar rice-based meals was lower on days when you walked within 30 minutes after eating.

Context

A fuller picture

Bring daily context into the glucose story

Individual readings are only one part of the picture. GlucoFinity is designed to organize the events around them so patterns are easier to inspect and discuss.

Continuous glucose trends

See daily patterns, time in range, and meal responses in one clear timeline.

AI-assisted meal recognition

Turn meal photos into reviewable food, serving, and carbohydrate estimates.

Prediction architecture

Prepare traceable estimated responses once a model has enough authorized data and measured evaluation.

Lifestyle pattern discovery

Compare glucose with meals, movement, sleep, and other daily context.

Sleep and activity context

Place morning and post-meal readings alongside rest and exercise events.

Medication logging

Record medication events as context without recommending dose changes.

Historical summaries

Review repeated responses and longer-term changes across days and weeks.

Privacy-focused handling

Designed around purposeful access, clear consent, and data minimization.

How it works

From scattered signals to reviewable context

Each part of the planned system has a focused role, from collecting user-approved data to presenting an understandable pattern summary.

  1. 01

    Connect health data

    Bring permitted glucose and health records into one timeline.

  2. 02

    Log meals and events

    Add meals, medication, sleep, activity, and useful notes.

  3. 03

    Analyze responses

    Compare timing, magnitude, and repeated glucose patterns.

  4. 04

    Discover patterns

    Review cautious summaries with the source context attached.

Daily context

CGM Meals Sleep Medication

Specialized analysis

Responses aligned by time and context

Understandable insight

A pattern summary with supporting context

Meal analysis demonstration

A meal estimate you can inspect and correct

The prototype proposes structured meal details and pairs them with the following glucose response. Every nutrition value remains an estimate until the user reviews it.

Meal photo placeholder

Recognized meal

Salmon rice bowl

12:35 PM - Fictional demo

AI-estimated nutrition

Brown rice

88% confidence

Estimated serving: 1 cup

45 g

estimated carbs

Grilled salmon

94% confidence

Estimated serving: 4 oz

0 g

estimated carbs

Roasted vegetables

82% confidence

Estimated serving: 1.5 cups

18 g

estimated carbs

Estimated total carbohydrates

63 g

Observed response

Three-hour post-meal curve

Peak of 168 mg/dL at approximately 60 minutes in this fictional example.

Demo response

Food recognition, serving sizes, nutrition values, and glucose predictions are estimates. Users should review meal details before using them as personal context.

Personalized insights

Patterns described with the right level of caution

GlucoFinity is designed to show what was observed, how much data supports it, and where uncertainty remains. These examples are fictional and informational.

Sample insight

Exercise timing

A post-meal walk was associated with a gentler rise

Across four similar rice-based meals, the observed peak was lower on days with a walk within 30 minutes.

Fictional evidence: 4 comparable meals observed

Sample insight

Sleep context

Shorter sleep may be related to higher morning readings

Morning glucose was higher after nights with less than six hours of logged sleep in this sample period.

Fictional evidence: 12 mornings compared

Sample insight

Meal variability

Similar meals produced different responses

A repeated pattern was observed: later dinners showed a wider response range than earlier versions of the meal.

Fictional evidence: 6 dinner responses reviewed

Sample insight

Repeated response

Oatmeal responses were relatively consistent

Three logged breakfasts had similar timing and peak ranges, with modest day-to-day variation.

Fictional evidence: 3 repeated breakfasts

Sample insight

Athletic context

Steadier pre-workout levels coincided with stronger sessions

Within the fictional training log, more stable glucose before exercise was associated with higher self-rated performance.

Fictional evidence: 8 training sessions logged

AI/ML foundation

Specialized models, clear responsibilities

The architecture separates meal interpretation, deterministic features, tabular prediction, time-series forecasting, statistical evidence, and explanation.

Status labels distinguish what this fictional demo calculates from foundational contracts and future models. No trained production model is represented as available.

01 / Inputs
Demonstrated

Normalized context

The demo preserves timestamps and mock-source provenance before calculations receive a reading.

  • Deterministic mock readings
  • Source-qualified records
02 / Meals
Foundation

Reviewable meal estimates

A provider-neutral contract returns structured food estimates while users retain control of every saved value.

  • Replaceable vision provider
  • Manual / AI / corrected provenance
03 / Features
Foundation

Tabular model pipeline

A deterministic feature version and chronological dataset rules prepare a future personalized XGBoost baseline.

  • Missing values remain missing
  • Versioned XGBoost evaluation
04 / Forecast
Planned

Continuous forecasting

Future sequence models remain separate from meal-response regression and require suitable longitudinal data.

  • 15–120 minute horizons
  • No neural model without data
05 / Explain
Foundation

Evidence before language

Statistics calculate supported associations first; a future language model may only explain supplied evidence.

  • Structured sample evidence
  • No invented medical conclusions

Safety and limitations

Designed to inform, never to replace care

Clear limits are part of a responsible health product, not fine print.

  • GlucoFinity is an educational and informational prototype, not a substitute for a licensed healthcare professional.
  • AI-generated insights may be incomplete, inaccurate, or based on insufficient context.
  • Nutrition values and glucose predictions are estimates that require user review.
  • Medication or insulin decisions should not be changed solely because of information shown by the application.
  • A production healthcare version would require rigorous validation, privacy protections, security controls, and regulatory review where applicable.

About the project

Built at the intersection of health and technology

GlucoFinity is an interdisciplinary university healthcare technology project exploring how complex glucose and lifestyle data could become more understandable without losing scientific caution.

The current site is a product concept and interactive demonstration. It does not imply university endorsement, clinical evidence, or a completed healthcare integration.

Computer science

Pharmacy knowledge

Bioengineering

Data science

Human-centered design

Interactive prototype

Explore the GlucoFinity Demo

See how fictional glucose, meal, sleep, activity, and medication data can be organized into a more understandable daily picture.

Explore the Demo Learn about the project

Current version: university project prototype using deterministic mock data.