lisTen.AI

Inspiration

We wanted to utilize real-time opinions to truly be valued, heard, and acted upon. Too many times network issues in certain areas or by certain groups of users go unnoticed. This hurts the workflows, timelines, and the reputation of users who depend on these networks to provide them with the essential tools to succeed in their education and profession.

What it does

We utilize live Reddit opinions to gauge customer happiness, current issues, and the spread of such issues. We then utilize a dashboard to display organized analytics to T-Mobile employees regarding the state of their service. In order to create organized analytics, an AI workflow is utilized.

How we built it

We utilized openrouter calls to NVIDIA's Nemotron model, Reddit's public API, a React frontend, a SupaBase (PostgreSQL) database, and Auth0 authentication to manage the overall functionality of the web app for AI-analytics, a modern UI, authentication/sign-in, data retention, data collection, data organization, and AI Agent workflows (a fetch agent to gather data from Reddit, an evaluator agent to organize the data in a meaningful manner and categorize it, and a report agent to generate natural language reports for employees to interact with.

Challenges we ran into

A lack of experience with database management and agentic workflows proved to be a challenge when it came to preserving data across sessions and designing a meaningful and coherent flow of information in our backend. However, a night of research and no sleep proved to be sufficient to tackle and overcome these issues.

Accomplishments that we're proud of

We are proud of having created a coherent backend flow and storage of data despite having almost no prior experience with tools/platforms such as Supabase and OpenRouter prior to the hackathon.

What we learned

We learned that despite seeming impossible at first, tasks are more feasible than they often seem. You will be surprised by what you can accomplish in 24 hours.

What's next for lisTen.AI

We would like to continue working on fetching data from better sources, such as X, since we were not able to gather enough data from data to provide truly accurate results. We would also like to improve the overall agentic workflow so that data is processed more accurately and add a fetch algorithm to improve the logic behind how often issues are diagnosed with compute resources in mind and to also truly automate the entire workflow. We would also like to improve the UI.