Pemmy

Inspiration

Project managers spend too much time on manual tasks like tracking updates, writing summaries, and finding information. We wanted to reduce this workload with an AI assistant that handles it automatically.

What it does

Pemmy is an AI chatbot that answers project questions, detects blockers and risks, creates action items, and summarizes meetings. It also displays analytics like progress, alerts, and team activity in one place.

How we built it

We used FastAPI for the backend, ChromaDB to store and retrieve project data, and NVIDIA’s Nemotron model to generate accurate, context-aware responses. The web interface was built using HTML, CSS, and JavaScript to support chat, actions, alerts, meetings, and analytics

Challenges we ran into

We faced issues when connecting to third-party tools, especially with setting up live data streams from platforms like Slack and Jira. Handling API authentication and permissions caused frequent errors, which made reliable data syncing difficult.

Accomplishments that we're proud of

Pemmy goes beyond a standard chatbot by proactively identifying issues, generating action items, and creating meeting summaries. These insights are organized and accessible through an integrated interface for easy tracking and follow-up.

What we learned

We learned how to work effectively with third party APIs and manage their data structures. We also gained experience setting up authentication flows, handling tokens securely, and ensuring smooth communication between services.

What's next for Pemmy

To scale Pemmy further, we plan to make setup and onboarding easier for new users. We also aim to add more third party data sources and improve our retrieval augmented generation system to deliver more accurate and reliable responses.