DevilNet: AI Productivity Assistant

An intelligent virtual assistant built with AWS Bedrock Agents, Lambda, and Google Workspace APIs (Gmail + Calendar) to help users manage tasks, schedule meetings, and stay productive all through NLP.

Comment

Inspiration

Managing emails, meetings, and tasks across tools like Gmail and Google Calendar can be overwhelming, especially when time zones and daily chaos get in the way. We wanted to create a truly intelligent virtual assistant that not only understands natural language but also acts on it — helping users stay productive without switching contexts.

What it does

AI Productivity Assistant is a voice/text-powered virtual assistant that integrates with Gmail and Google Calendar to help users:

📅 Schedule, reschedule, or check calendar events

📧 Read, prioritize, and send emails

🔍 Interpret natural language like “next Monday” or “in 3 hours”

🧠 Ask clarifying questions to improve task understanding

⏰ Automatically adapt to the user’s timezone (default: UTC-7)

Example prompts:

“What meetings do I have today?”

“Send a thank-you email to Alex for the presentation.”

“Schedule a call with the design team tomorrow at 11 AM.”

How we built it

AWS Bedrock Agents: Powers the core language understanding and reasoning

Google Workspace APIs: Used for reading/sending emails and managing calendar events

Action Groups: Serve as Bedrock’s bridge to Gmail and Calendar actions (read_gmail, send_gmail, create_calendar_event, etc.)

AWS Lambda: Executes secured logic when actions are triggered

Python: Handles authentication and communication with Google APIs

Challenges we ran into

Authenticating securely with Gmail and Calendar APIs through Lambda

Dynamically interpreting vague natural language times (e.g. “next Friday morning”)

Maintaining context across multiple user inputs without re-authentication

Ensuring real-time responsiveness while handling async workflows

Accomplishments that we're proud of

Seamlessly combining AWS Bedrock Agents with real-world APIs for productive outcomes

Built a working, natural-language interface to both Gmail and Google Calendar

Enabled context-based follow-up questions for more intuitive interaction

Delivered timezone-sensitive scheduling without user having to specify it

What we learned

How to design action groups for Bedrock Agents

Interfacing Google Workspace APIs with AWS Lambda securely

Handling fuzzy time parsing and context tracking

The power (and challenge) of creating intelligent assistants that feel human

What's next for AI Productivity Assistant

🗣️ Voice command integration via Alexa or WebRTC

🔐 OAuth-based user-specific login flow for calendar/email access

📊 Daily productivity summaries and reminders

🧩 Integrations with other productivity tools like Notion, Slack, and Trello

🧠 Long-term memory and preferences for personalized assistance

Built With

Share this project:

Updates