Caresync

Use LLMs and AI agents to automate surgery planning by creating tasks, booking the OR, assigning tasks based on staff availability, and handling all planning steps for the doctor.

Comment

🧩 Product & Design Specification: careSync_

📌 Project Overview careSync_ is an AI-powered surgical planning and hospital coordination platform that automates and streamlines the pre-operative scheduling process. It integrates a fine-tuned LLM for surgical planning, resource availability checks via backend agents, and a task management system for doctors and hospital staff.

🎯 Objective To reduce the overhead and delays in surgical case planning by:

  • Automatically generating surgical workflows using a fine-tuned LLM
  • Dynamically querying the availability of staff, operating rooms, nurses, and equipment
  • Allowing doctors to review, publish, and manage schedules and tasks from a single interface

🏗️ System Architecture (High-Level) Flow:

  1. The doctor gets case input via the EHR.
  2. A fine-tuned LLM (Phi-2 + LoRA) interprets the case and generates a structured surgical plan (in JSON format).
  3. A backend agent crew (using LangChain) processes this plan by:
    • Checking resource availability (doctors, nurses, ORs, equipment)
    • Matching schedule constraints
  4. The draft plan is returned to the frontend for review.
  5. Upon approval, the doctor clicks Publish, which:
    • Sends confirmation emails or logs to assigned personnel
    • Finalizes the schedule in the database
  6. Doctors can create, update, and remove additional tasks related to the patient.

A supporting architecture diagram is attached in the final submission.

🧠 Core Components Layer | Description ------------|------------------------------------------------------------ LLM Engine | Fine-tuned Phi-2 model using LoRA for medical instruction Backend | FastAPI server using LangChain + MongoDB for agent execution Frontend | React + Tailwind dashboard for doctors Database | MongoDB database stores resources, availability, tasks

🗃️ MongoDB Collections Used

  • staff
  • equipment_availability
  • nurse_availability
  • ot_availability
  • test_history
  • tasks

📈 Data Flow Summary Doctor gets case → LLM generates structured plan → Backend agents resolve availability → Draft returned → Doctor approves and publishes → Agents notify and log → Doctor creates follow-up tasks → All saved in DB

📊 Frontend Workflow (Doctor Experience)

  1. Doctor clicks AI Suggestion → triggers LLM and backend
  2. Views the generated surgical plan with required resources
  3. Confirms plan via Publish → sends notifications, books OT
  4. Manages tasks through Create Task, Update Task, or Remove Task

🧪 Key API Overview

  • POST /availability – Checks the availability of required hospital resources
  • POST /publish – Finalizes and notifies stakeholders of surgical schedule
  • POST /tasks, GET, PUT, DELETE – Full CRUD for task management

🚀 Future Enhancements

  • Google/Outlook calendar integration for real-time sync
  • Role-based access (doctor, nurse, admin)
  • Task history and audit logs
  • Attachments (PDF orders, reports)

📬 Contact & Team Developed during Commure Codes Hackathon by:

  • Jayesh Shinde – Backend, LLM, API architecture
  • Gaurav Patil – Agents, MongoDB, Task System
  • Maulik Ghatala – Frontend, UI/UX

Built With

Share this project:

Updates