CS 491/492 Senior Design Project · Bilkent University
DiaComp
A calm digital companion that reduces the mental burden of living with Type 1 diabetes. It remembers the user's own past episodes and speaks only when it is likely to help.
Project Overview
DiaComp is a calm digital companion designed to reduce the mental burden of living with Type 1 diabetes. By continuously observing glucose data and learning from a user's past experiences, it aims to identify moments when timely support may be useful without adding to the constant stream of alarms and notifications already present in diabetes management.
Rather than replacing existing medical devices or treatment plans, DiaComp focuses on the overlooked emotional and cognitive side of the condition: helping users feel less alone and less anxious. Its core innovation is combining deterministic safety rules with personalized historical context and carefully constrained language-model-generated guidance to provide support only when it is genuinely needed.
Why This Matters
One of us has Type 1 diabetes and uses a continuous glucose monitor, a cordless insulin pump and an artificial pancreas algorithm. Blood sugar control is good, yet none of this equipment does anything about the mental side of the condition.
Compared to people without Type 1 diabetes, people with it have:
the chance of clinical depression
the risk of suicide
the chance of an eating disorder
These are not caused by poor blood sugar control. They come with the condition itself.
What Current Technology Covers
| Device | Solves | Does not solve |
|---|---|---|
| Glucose monitor | Knowing the number | Having to keep checking it |
| Insulin pump | Delivering insulin accurately | Deciding when to act |
| Artificial pancreas | Holding blood sugar in range | Fear of the times it fails |
| Alarms and reports | Flagging a limit, summarising a week | How the day felt |
DiaComp works on the right-hand column.
What We Target
Loneliness
Decisions are made alone. There is usually nobody nearby who understands the situation, and the decisions keep coming.
Constant anxiety
A background fear of the next low or high, which does not stop, because the next one is always possible.
This is our reading of the cause, not a proven mechanism. Neither is helped by better insulin delivery; both can be helped by software that shows up at the right moment and knows the user's history.
Key Ideas
Episodic memory
Other apps store a list of glucose numbers. DiaComp stores episodes: the situation, what it did, and what happened next.
Silent by default
Fixed algorithms decide whether a moment is worth a message. Most of the time, the answer is no.
The LLM writes, never decides
A language model phrases the message. Rules decide whether it is sent, and a safety layer checks it before display.
Safety in code
Limits are written in code, not in prompts, and covered by 380 automated tests.