• 

🏥 Health Recovery Bot Using LLMs 🤖

🔍 Introduction

Recovering from an illness or injury can be confusing and overwhelming for patients. Many struggle to understand:

❓ What exactly am I suffering from?

❓ How serious is my condition?

❓ What can I do to recover faster?

❓ What should I avoid?

🚑 Lack of clear, reliable information can lead to:

1. Delayed recovery due to poor self-care.

2. Unnecessary anxiety about symptoms.

3. Worsening health due to incorrect assumptions.

4. To solve this, we propose the Health Recovery Bot, an AI-powered assistant that helps patients understand their condition, track progress, and receive actionable recovery guidance.

⚠️ The Problem

🩺 Patients Often Feel Lost During Recovery Many patients lack clarity about their condition and recovery process: 🔹 48% of patients don’t fully understand their diagnosis or treatment plans. (NCBI)

🔹 30% of hospital readmissions are due to patients misunderstanding aftercare instructions. (AHRQ)

🔹 60% of people turn to the internet for health advice, often finding misleading or harmful information. (Pew Research)

🚑 Consequences of Poor Understanding

⚠️ Slower Recovery – Not following proper aftercare delays healing.

⚠️ Health Risks – Misinterpreting symptoms can lead to serious complications.

⚠️ Increased Anxiety – Patients often panic over normal recovery symptoms.

⚠️ Over-reliance on Doctors – Many unnecessary visits could be avoided with better guidance.

⚠️ Confusion from Generic Advice – Many health articles online are too broad and may not apply to a specific case.

⚠️ Difficulty in Identifying Progress – Patients struggle to know if they are improving or if they need medical attention.

🏥 Current Challenges in Patient Support

💬 Doctors Have Limited Time – Physicians cannot answer all patient questions in detail.

📄 Discharge Papers Are Overwhelming – Many patients find hospital instructions too technical or hard to follow.

🌍 Healthcare Gaps Exist – In remote areas or developing regions, patients may have limited access to doctors after discharge.

🤖 Most Chatbots Are Too Generic – Existing healthcare bots lack personalization and do not adapt to the user’s specific case.

💡 The Solution: Health Recovery Bot 🤖

The Health Recovery Bot is a personal AI assistant that helps patients understand their recovery journey with clear, personalized, and reliable information.

🚀 How It Works

1️⃣ User Inputs Symptoms & Questions

The patient upload thier medical reports

2️⃣ AI Analyzes & Retrieves Trusted Medical Knowledge

Uses Mistral LLM to interpret symptoms and questions. Uses RAG (Retrieval-Augmented Generation) to provide fact-based responses.

3️⃣ AI Provides Personalized Recovery Guidance (Couldn't build in the time limit) Explains what the condition means in simple language. Tells how severe it is and what to expect during recovery. Suggests do’s and don’ts based on medical best practices.

🛠️ Technical Implementation

🔹 Flask – Web framework for API & chatbot interface.

🔹 Ollama + Mistral – LLMs to generate accurate & human-like responses.

🔹 Retrieval-Augmented Generation (RAG) – Ensures answers are based on trusted medical sources.

🔹 Chroma DB – Stores and retrieves patient interactions & knowledge efficiently.

🎯 Conclusion

The Health Recovery Bot will empower patients by providing clarity, confidence, and control over their recovery.

✅ Faster recovery with better self-care.

✅ Reduced anxiety through clear explanations.

✅ Fewer unnecessary doctor visits.

✅ Trustworthy AI-powered support.

🚀 Let’s transform patient recovery with AI!

Inspiration

Recently one of my close relatives was suffering from Kidney stones and they thought it was happening due to the presence of stones around them. Well even though this is one of the extreme cases, we all often misunderstand our ailment and I really wanted to make something that explains and supports us during the recovery process.

What it does

I am trying to build a solution that uses LLM to break down our medical reports, explain to us, and further answer FAQ regarding what we should do and avoid during the recovery process.

How we built it

The project is built in a flask with tailwind UI. It takes in PDF uploads it to the server, and uses Chroma vector to create context out of it. That context is then fed to QA chain in ollama which uses Mistral as LLM.

Challenges we ran into

TIME!! And getting the running Query chain running.

Accomplishments that we're proud of

I am submitting!!!!

What we learned

Fixing bugs sucks a lotttttttttt in time pressure.

What's next for Recovery assistant

I was able to make a very very small fraction of the project during the given time given that I had to start the project 3 times from scratch in different languages. I hope to continue with Python and complete its mission

Built With

• flask
• mistral
• ollama