MeetMind

Inspiration

I have too much Microsoft Teams meetings and I wanted to have a way such that I have my own secretary being able to process/ingest information, and provide additional insights

What it does

This agent is suppose to be able to ingest video data and transform it into Notes, especially with the full integration of Github, Jira, as well as Confluence Wiki.... (Latter two hasnt been done yet)

How we built it

We used Railtracks to create the base agentic framework, senso to provide lasting context to the agents, and Digital Ocean's serverless inference to provide the LLM inference required; We do, however, have to use our own openRouter to ingest Gemini 2.5 flash models to provide the visual/audio data

Challenges we ran into

It is quite challenging as we dont have alot of actionable data + I do not have another individual to help build this with. The personas needs to be more diversified while i only have SWE data... ALSO we need more different personas to train/finetune the prompt and build a better harness around it

Accomplishments that we're proud of

We got a working demo for the first time within 5 hours, and that is something we should be proud of end-2-end

What we learned

Building App is the easy part; getting the appropriate data would be infinitely difficult as now the product requires data to be fed back in to become a better product: I need to also integrate more on the visual aspect of this project

What's next for MeetMind

Context-aware meeting notes with timestamps integration with Jira, Github, and Confluence Wiki for maximum knowledge base parse through Perhaps integrating this with SEMIONT database to have agents be able to use meeting nots as a source of truth!

Built With

• do
• opencode
• python
• railtracks