FDABot Application

Inspiration

The inspiration behind FDABot stemmed from the need for a reliable, accessible, and user-friendly tool that could provide up-to-date information on FDA enforcement actions. Investment and healthcare professionals often require this information as it can significantly impact markets or public health. We wanted to create a solution that not only meets this need but also adds an element of fun by allowing users to verify rumors, check on colleagues, or find specific enforcement actions easily.

What it does

FDABot is an AI-powered application that uses the FDA's enforcement actions database as its knowledge base. It serves as a valuable resource for investment and healthcare professionals who need timely and accurate information about FDA enforcement actions. Users can quickly verify rumors, check if colleagues are under audit, or search for specific enforcement actions. The application offers a seamless and engaging way to access critical information that could influence market decisions or public health strategies.

How we built it

FDABot was built using a combination of advanced AI and natural language processing (NLP) technologies. We integrated the FDA's enforcement actions database into the AI model to ensure comprehensive and up-to-date information. The development process involved:

1. Data Integration: Collecting and integrating data from the FDA's enforcement actions database using google/puppetteer and snowflake's embeddings model (large) .
2. AI Model: we're using snowflake's arctic model (instruct)
3. User Interface Design: obviously using streamlit and their cloud hosting service (which is very convenient!)
4. Putting it Together: We're using llamaindex integrations and frameworks to orchestrate the application.

Challenges we ran into

During the development of FDABot, we encountered several challenges:

1. Data Accuracy: Ensuring the data from the FDA's enforcement actions database was accurate and up-to-date.
2. AI Training: Training the AI model to understand and accurately process a wide range of user queries.
3. User Interface: Designing an interface that is both intuitive and functional for diverse user groups, including investment and healthcare professionals.
4. Performance Optimization: Ensuring the application runs smoothly and provides quick responses despite the large volume of data.

Accomplishments that we're proud of

We are proud of several key accomplishments with FDABot:

1. Accurate AI Model: Successfully training an AI model that can accurately interpret and respond to user queries about FDA enforcement actions.
2. User-Friendly Design: Creating an intuitive and engaging user interface that enhances the user experience.
3. Real-Time Information: Providing users with up-to-date and reliable information that is crucial for their professional needs.
4. Positive User Feedback: Receiving positive feedback from beta testers, which validated the effectiveness and usability of the application.

What we learned

Throughout the development of FDABot, we gained valuable insights:

1. Importance of Data Quality: The critical role of accurate and up-to-date data in building a reliable AI application.
2. User-Centric Design: The necessity of designing with the end-user in mind to ensure the application meets their needs and preferences.
3. Continuous Improvement: The value of iterative development and continuous feedback to enhance the application's performance and user experience.
4. Collaboration: The benefits of cross-disciplinary collaboration in developing a robust and functional AI application.

What's next for Team Tonic

Looking ahead, Team Tonic has several plans for FDABot:

1. Feature Expansion: Adding new features to enhance functionality, such as personalized alerts and advanced search options.
2. Broader Data Integration: Integrating additional data sources to provide a more comprehensive knowledge base.
3. User Community Engagement: Building a community of users to foster engagement and gather ongoing feedback for improvements.
4. Scaling Up: Expanding the application's reach to more professionals in the investment and healthcare sectors, and exploring potential applications in other industries.

We are excited about the future of FDABot and are committed to continually improving and expanding its capabilities to better serve our users.

Built With

• arctic
• embedding
• fda
• huggingface
• json
• llama-index
• puppetteer
• python
• snowflake
• streamlit