AI Series: How Haema is using AI in healthcare

In our new AI series section, we are writing about startups that are innovating in the space of health and using AI and LLMs. In today’s feature, we spoke with Jishma and Pranav, the cofounders of Haema, who are creating an AI-based diabetes assistant.

The Inspiration 

Pranav and Jishma met during their university years in Bath. Pranav was a robotics student, while Jishma pursued pharmacology. Their academic paths might have seemed divergent, but a personal challenge brought them together on a shared mission. Jishma had been living with type 1 diabetes since she was two years old. Managing her condition was a daily struggle, involving constant monitoring and meticulous planning to keep her blood sugar levels stable. Pranav, fascinated by the potential of technology, was determined to help his friend find a better way.

Throughout their university years, Pranav and Jishma brainstormed and experimented with various methods to ease the burden of diabetes management. They tracked blood sugar levels, documented meals, and analyzed the impact of different foods and activities. It wasn't just about collecting data; it was about understanding patterns and making life a bit easier. After graduation, equipped with more knowledge and experience, they decided to turn their experiments into something tangible. This was the beginning of Haema.

For Jishma, Haema was more than a project; it was a lifeline. Growing up, she had to inject herself with insulin multiple times a day, manually calculate carbohydrate intake, and use finger-prick tests to monitor her blood sugar. This process was time-consuming and often painful. Over the years, advancements in technology brought significant improvements. Jishma transitioned from insulin injections to using an insulin pump, a device that automatically delivers insulin into her body based on her carbohydrate intake. Additionally, the development of continuous glucose monitors (CGMs) allowed her to constantly monitor her blood sugar levels without the need for frequent finger-pricks.

Revolutionising Diabetes Care: The First-Ever AI Diabetes Assistant

Despite these advancements, managing diabetes still required significant effort and vigilance. Haema took these improvements to the next level.With Haema, Jishma can simply take a picture of her meal and get an instant analysis of its nutritional content and its potential impact on her blood sugar levels. This technology reduces the need for manual calculations and constant monitoring, making diabetes management more manageable and less intrusive in her daily life.

Using LLMs since the beginning

The initial version of Haema was simple. It started as a website, featuring a large language model to interact with users and access six months of Jishma's blood sugar data. This proof of concept allowed users to log their meals and receive feedback, laying the foundation for a more comprehensive tool. Realizing the potential, Pranav and Jishma decided to commit fully to their project. They moved to San Francisco for a few months, immersing themselves in the startup culture and refining their app.

Evolution Through Iteration

The development of Haema is a journey of constant iteration and improvement. User feedback is invaluable, guiding Pranav and Jishma as they enhance the app’s functionalities. Haema evolves from a basic logging tool to an AI-powered mobile application capable of analysing meals, predicting blood sugar impacts, and offering personalised recommendations. The app integrates data from glucose monitors, exercise trackers, and other wearables, providing a holistic view of the user’s health.

AI at the Core

Haema's core technology relies particularly on large language models and plays a crucial role in multiple aspects of the app's functionality:

1. Meal Analysis: Users take a picture of their meal, and the AI identifies the ingredients, estimates the macronutrient content, and predicts how the meal will affect blood sugar levels. This analysis is grounded in a comprehensive database of food research.

2. Personalized Recommendations: The AI provides instant recommendations on how to adjust meals to minimise blood sugar spikes. For example, it might suggest adding more protein to a meal to balance blood sugar levels.

3. Behavioral Insights: By analysing data from continuous glucose monitors (CGMs) and other wearables, the AI helps users understand how different activities—like exercise or sleep—affect their blood sugar. This allows for personalised advice tailored to each user’s unique physiological responses.

AI's Role in Health

Haema’s AI-driven approach extends beyond meal analysis. It aims to be a personal health companion, helping users set and track health goals. The app provides tailored advice based on patterns in the user’s data, such as suggesting dietary adjustments or optimal exercise times to maintain stable blood sugar levels. The AI’s ability to learn and adapt to each user’s unique needs makes Haema an invaluable tool for diabetes management.

The AI health advisor

Pranav and Jishma envision Haema as a personal AI health advisor for everyone, not just those with diabetes. They aim to help people manage chronic illnesses, achieve fitness goals, and maintain overall well-being. Their long-term goal is to reach the millions of people worldwide living with diabetes, providing them with a tool that can significantly improve their quality of life.