Inspiration

Recent advancements in AI and generative models have been a major inspiration for creating our product. With models like GPT-4 and Gemini Pro showing unprecedented capabilities in natural language understanding, reasoning, and multimodal processing, we saw the potential to revolutionize healthcare diagnostics. These technologies demonstrated how AI can generate accurate, context-aware insights in real time, which sparked the idea of applying this power to critical, high-stakes environments like the ER. Our product leverages these trends by combining AI’s ability to process vast medical data, real-time vitals, and imaging to assist doctors in making faster, more informed decisions.

What it does

Our AI assistant seamlessly integrates with existing hospital systems, including EPIC, to provide real-time diagnostic reasoning for critical ER cases.

It processes patient data from multiple sources (e.g., vitals, imaging, labs, doctor’s notes) and generates rapid, evidence-based insights, assisting doctors in making timely and accurate decisions.

How we built it

We used a hierarchical, multi-agentic framework, to process different modalities of ER data such as time series vitals, CT/MRI imaging, doctor and nurses notes to build a real-time ai-based system that can help warn against early indicators of sepsis

Challenges we ran into

  • very slow wifi at the venue
  • inference issues with large language models
  • lack of time management

Accomplishments that we're proud of

  • Made new friends and worked on an inspiring new idea together

What we learned

  • It's important to prototype more quickly

What's next for Pognosys

  • Research into avenues of Neuro Symbolic AI to create novel approaches for reasoning capabilities of LLMs in the medical space

Built With

  • geminipro
  • kindo
  • langchain
  • python
  • streamlit
Share this project:

Updates