Robin

Automate busy work by talking to Robin, an AI agent that interacts with your workspace on your behalf, handling emails, chats, and calendar tasks for smarter, faster productivity at scale

Comment

Inspiration

Robin was inspired by the possibility of getting real work done hands-free, wherever you are. You simply talk to your trusted agent and it organizes your day, takes care of busy work, and keeps you informed, so you can jump right into important tasks wherever you are. Robin helps you focus on what matters most and makes sure you're always ready for action.

What it does

Automate busy work just by talking to Robin. The agent interacts with your workspace on your behalf, managing your email, calendar, Slack, and more (coming soon). Robin listens to your commands, handles routine tasks, coordinates your schedule, and keeps you updated, so you are always ready for what comes next.

How we built it

Tech Stack:

  • Amazon Bedrock (Nova)
  • AgentCore runtime/gateway
  • React Native (iOS/Android)
  • Python backend
  • Custom WebSocket service

Integrations:

  • Gmail
  • Google Calendar
  • Slack
  • Modular architecture for both cloud and mobile, running on AWS
  • Dockerized deployments and Amazon ECR for region flexibility

Challenges we ran into

  • Migrating infrastructure across AWS regions and rewriting components for region-specific services
  • Keeping latency low
  • Writing the system prompts
  • Runtime failing to deploy on multiple occasions (quota problems)

Accomplishments that we're proud of

  • Built an autonomous agent that actively handles office work and makes mornings more productive

What we learned

  • Advanced deployment and scaling with AWS (Bedrock, ECR, AgentCore)
  • Patterns for building multi-service, voice-driven agent experiences
  • Different methods agents can complete meaningful actions in the real-world

What's next for Robin

  • Expanding connections to more apps like Asana, Jira, Confluence, and other tools
  • Smarter agents that can summarize, suggest actions, and learn from your preferences
  • Personalities for one-trick pony agents

Built With

Share this project:

Updates