AI Engineer Fresher

Modern organizations receive hundreds (sometimes thousands) of emails daily. Many of these emails are support-related (e.g., customer queries, requests, or help tickets). Manually sifting through these emails, prioritizing them, and drafting professional responses can be time-consuming and error-prone.
Your challenge is to build an AI-Powered Communication Assistant that can intelligently manage these emails end-to-end. The assistant should analyze incoming emails, prioritize them based on urgency, generate appropriate responses, and display the results on a user-friendly dashboard.
The goal is to improve efficiency, response quality, and customer satisfaction while reducing manual effort.

Core Requirements:

• Email Retrieval & Filtering
  • Fetch all incoming emails from the user’s email account (IMAP/Gmail/Outlook APIs, etc.).
  • Filter emails whose subject lines contain terms like:
    • “Support”
    • “Query”
    • “Request”
    • “Help”
  • Extract and display relevant details for each filtered email on the dashboard:
  • Sender’s email address
  • Subject
  • Email body
  • Date/time received
• Categorization & Prioritization
  • Each eligible email should be automatically categorized and ranked:
    • Sentiment Analysis: Positive / Negative / Neutral
    • Priority: Urgent / Not urgent (e.g., based on keywords like “immediately,” “critical,” “cannot access,” etc.)
    • Emails marked as urgent should appear at the top of the list for processing (implement a priority queue for email processing) 
• Context-Aware Auto-Responses
  • Use an LLM (Large Language Model) to generate draft replies for each incoming email.
  • Responses must:
    • Maintain a professional and friendly tone
    • Be context-aware (e.g., should use a knowledge base to answer these questions (RAG + Prompt Engineering) + respond contexually - if the customer is frustrated, acknowledge their frustration empathetically)
    • Include relevant details (e.g., if the email mentions a product, reference it in the reply)
    • Prioritized emails (urgent) should be responded to first.
• Information Extraction
  • From each incoming email, extract key information such as:
    • Contact details (phone number, alternate email, etc.)
    • Customer requirements or requests
    • Sentiment indicators (positive/negative words)
    • Any metadata that can help support teams act faster
    • This information should be displayed clearly on the dashboard, alongside the raw email.
• Dashboard / User Interface
  • A simple, clean dashboard where:
    • The Filtered support emails are listed along with extracted key details in a structured format
    • Anaytics and Stats are present - Categories (sentiment + priority), total emails recieved in last 24 hours, emails resolved, emails pending, interactive graph to display these details
    • AI-generated responses are displayed (and can be reviewed/edited before sending)

Technical Guidelines:

• You can use any tech stack (Python/Node.js/Java + React/Next.js for frontend, etc.).
• For (sentiment, summarization, response generation), you may use: OpenAI GPT models, Hugging Face models (BERT, T5, DistilBERT, etc.)
• For email retrieval, use: Gmail, Custom SMTP, Outlook Graph API, or standard IMAP libraries.
• For storage, a simple database (SQLite, MongoDB, or PostgreSQL) is sufficient.

Evaluation Criteria

• Functionality: Accuracy: How well the Email functionalities like reading emails (based on filters) and writing emails/auto-sending of replies (based on priority). Does the assistant correctly filter, categorize, and contextually respond to emails based on priority

• User Experience: Is the dashboard intuitive and contains all the information to give users the right solution. 
• Response Quality: How accurately the AI techniques, Retrieval-Augmented-Generation (RAG), Prompt Engineering and Context embedding, are applied to generate correct response. 

Timeline:

Hackathon Duration: 4 days

Deliverables:

• End-to-end working solution of AI-Powered Communication Assistant
• Demonstration video of all the working features of the platform
• Short self-written documentation (no AI) on architecture + approach used

Impact:

• This assistant has the potential to transform customer support operations by:
• Reducing manual workload
• Ensuring faster, empathetic responses
• Extracting actionable insights from unstructured emails
• Improving customer satisfaction and retention