Aidol

Inspiration

Last year, I experienced a severe case of cholecystitis, which was both frightening and challenging. When I arrived at the hospital, I found myself struggling to articulate the nature of my pain and unsure about which clinic would be most appropriate for my condition. After enduring two long hours of examinations and probing questions, I came perilously close to a life-threatening situation. This harrowing experience is what motivated us to create this innovative solution.

What it does

Our tool empowers users to describe their symptoms with greater accuracy and detail. By doing so, it significantly enhances the efficiency with which doctors can assess and diagnose medical conditions. This streamlined communication between patients and healthcare providers not only saves valuable time but can also lead to faster and more appropriate treatment.

How we built it

To tackle the issue of misinformation and ambiguity in symptom description, we implemented a systematic approach by categorizing various hospital outpatient departments. This strategy helped us minimize instances of hallucination, allowing for clearer communication and more relevant responses based on patient input.

Challenges we ran into

During the process of selecting appropriate departments, we encountered challenges where the language model (LLM) could generate specific analyses but often fell short in providing concrete choices. This inconsistency made it difficult to ensure that users received the most relevant and actionable information when they needed it most.

Accomplishments that we're proud of

One of our significant achievements in this project has been the successful development and implementation of our proprietary multi-agent programming framework, known as nerif. This framework has proven to be instrumental in enhancing the operational capabilities of our tool, allowing for a more refined and efficient user experience.

What we learned

Through our experiences, we discovered that leveraging a multi-agent framework substantially lowers the complexity involved in API development. Moreover, it facilitates more accurate guidance for users by utilizing log probabilities within the API, resulting in a more informative and user-friendly interaction.

What's next for Aidol

Looking ahead, we have ambitious plans for the continued evolution of Aidol. Our next steps include:

1. Expanding post-training efforts to incorporate a Chain of Thought (COT) methodology, which will enable us to prompt users to provide more detailed and pertinent information about their symptoms.
2. Developing multimodal capabilities that will allow us to transform the final generated reports into audio recordings, thereby enhancing accessibility and user engagement with our platform.

Built With

• nerif
• python