Inspiration
Modern teams spend hours in meetings, yet execution still breaks down. Action items are missed, ownership is unclear, and follow-ups rely on manual effort. We were inspired to build an AI that doesn’t just summarize conversations, but actively owns execution like a real Product Manager.
What it does
AURAS AI is an autonomous AI Product Manager that converts meetings into structured execution. It transcribes discussions, extracts tasks, assigns ownership, detects blockers over time, and proactively nudges teams before delays compound.
How we built it
We built AURAS using Gemini 3 API as the core reasoning engine. Gemini 3 processes meeting transcripts, reasons over long-term context, and generates explainable task decisions. Outputs are validated against schemas, logged with confidence scores, and surfaced through a human-in-the-loop task review system.
Challenges we ran into
The biggest challenge was trust. AI systems often hallucinate or overcommit. We solved this by prioritizing precision, adding explicit uncertainty handling, and requiring human approval for high-impact actions.
Accomplishments that we're proud of
- End-to-end meeting → task → follow-up automation
- Blocker detection across historical meetings
- Explainable AI decisions with reasoning and confidence
- Privacy-first design with data isolation and redaction modes
What we learned
Trust in AI comes from transparency, not confidence. Users adopt AI faster when it explains its actions and allows control rather than forcing automation.
What's next for Auras AI
We plan to expand integrations (Jira, Slack, Teams), improve execution forecasting, and evolve AURAS into a full execution intelligence layer for modern teams.
Built With
- caching
- chat
- docker
- docker-compose
- express.js
- fastapi
- fastapi-ai-&-reasoning:-gemini-3
- gemini-3-(task-extraction
- gemini3
- integration-&-workflow-tests-apis:-restful-apis-for-ai-analysis
- jest
- jira
- jwt
- langchain
- langchain-(agent-orchestration)-speech-&-nlp:-whisper-(speech-to-text)
- mongodb
- mongodb-caching-&-queues:-redis-(job-queues
- multi-agent-coordination)-integrations:-jira-api
- netlify-testing:-jest
- next.js
- node.js
- notion
- notion-api-(slack-&-teams-planned)-auth-&-security:-jwt-based-auth
- postgresql
- prioritization
- privacy-mode-data-isolation-infrastructure-&-devops:-docker
- python
- python-frontend:-react
- react
- reasoning
- redis
- render
- restfulapi
- role-based-access-control
- slack
- summarization)-databases:-postgresql-(with-pgvector)
- tailwind
- tailwind-css-backend:-node.js
- task-extraction
- typescript
- whisper
Log in or sign up for Devpost to join the conversation.