ParaDoc

Visualize your health decisions before making them , ParaDoc simulates real outcomes so you can choose smarter.

Comment

🧠 Inspiration

In moments of health uncertainty , a sudden chest pain, a lingering cough , most people face a terrifying decision: wait, Google, or panic. We wanted to build something that offered clarity, calm, and medically-informed direction.

That’s how ParaDoc was born , a platform that helps people simulate future outcomes of health decisions through AI-powered analysis and branching scenarios.

We asked:
What if you could simulate how your condition might progress depending on what you do next , like ignoring it, resting, or seeking medical help , and see a clear breakdown of how that choice might evolve over 7 days?

⚙️ What It Does

ParaDoc allows users to input symptoms and instantly see multiple possible response paths:

  • Doing Nothing
  • Self-Treatment
  • Seeking Medical Attention

For each path, ParaDoc provides a 7-day simulated timeline showing likely events day-by-day (e.g. symptom progression, treatment milestones, recovery signs). It also gives:

  • A Risk Score (likelihood of complication)
  • A Recovery Score (probability of improvement)
  • A recommended best path based on real-world context

This lets users make informed decisions based not on panic, but personalized insight.

🛠️ How We Built It

  • Frontend: Built with Next.js and Tailwind CSS for responsive and intuitive UI
  • LLM Integration: GPT-4o and Gemini power timeline generation and scoring using structured prompts
  • Prompt Engineering: Custom modular prompts to simulate diverse symptoms and scenarios with minimal repetition or bias
  • Pathway Logic: Backend logic generates 5-7 timelines per input, each with its own stats, summary, and outcome scoring

🚧 Challenges We Faced

  • Designing prompts that created distinct but realistic timelines
  • Ensuring the scoring wasn’t biased toward only one path (e.g. always choosing "doctor visit")
  • Balancing clinical realism with approachable language
  • Handling dynamic symptom input while keeping the output coherent and medically consistent

💡 What We Learned

  • People engage deeply with branching narratives when it comes to health
  • Even simple visual cues (risk/recovery score, step timeline) help people trust AI decisions more
  • Language and UX matter as much as accuracy when the goal is helping someone in distress

🌱 What’s Next

  • Add comic-style timeline generation using Stable Diffusion
  • Integrate voice input and audio narration for accessibility
  • Localize for multilingual rural populations
  • Build a clinician-facing dashboard to create verified simulation templates
  • Launch a mobile-first version for easy access on low-bandwidth networks

Built With

Share this project:

Updates