ResuRalph 🤖📄

A Discord bot for collaborative resume feedback

In today's competitive tech job market, a standout resume is the first step toward success. However, the resume review process can often be cumbersome.

• Long, overwhelming review threads, where comments can easily get buried and overlooked.
• Reviewers forced to download PDFs to view resume's and updated resume's, which becomes increasingly tedious with each incremental change.
• The frustration of manually identifying and specifying which parts of the resume you're referring to.
• Comparing two resume PDF's and trying to identify what changes were made by the user.

ResuRalph streamlines this process.

Demo 🎬

The Flow ⏳

Initial Upload

• Upload your resume as a PDF using the /upload command.
• ResuRalph validates the PDF, stores it in AWS S3, and saves metadata in DynamoDB.
• The bot generates a Hypothes.is annotation link for collaborative review and sends it back to the Discord thread.
• With this link, reviewers can leave in-line annotations directly on your resume.

Getting Feedback

• Use /get_annotations to pull annotations from reviewers directly into Discord.
• Try /ai_review for AI-powered feedback and annotations (limited to once per day).
• View all your resume versions with /get_all_resumes.
• Get your latest resume link anytime with /get_latest_resume.

Updating & Comparing

• Use /update to upload an improved version of your resume.
• Enable the show_diff option to see changes between versions: 🟢Added: "Project X | React, Node, SQL..." 🔴Removed: "Work Experience Y | Example Company..."
• Use /get_resume_diff to compare any two resume versions.
• /clear_resumes permanently removes all your data from the system.

---

Tech Stack 🛠️

Backend Architecture

• Python 3.13 - Core application language
• Flask + Mangum - Web framework with AWS Lambda adapter
• Discord Interactions - Discord bot integration
• Docker - Containerized deployment

AWS Infrastructure

• AWS Lambda - Serverless compute (ARM64 architecture)
  • Main Lambda: Discord interaction handler (10s timeout)
  • Command Processor Lambda: Async command processing (60s timeout)
• Amazon S3 - PDF storage with public access
• DynamoDB - Resume metadata and user data storage
• SQS - Asynchronous command processing queue
• CDK (TypeScript) - Infrastructure as Code

---

A Challenge: Discord's 3-Second Limit

Discord requires bot responses within 3 seconds, but operations like AI reviews and PDF-diff (for resume updates) can take 10-ish seconds. These approaches failed:

• Background threads in Lambda: Lambda terminates background threads when the main handler returns
• Thread joining: Blocks the response, violating the 3-second limit
• Synchronous processing: Times out and fails user interactions

Solution: Deferred Response with SQS + Seperate Command Handler Lambda

1. Immediate Response (< 3 seconds)

   • Main Lambda receives Discord interaction
   • Returns deferred response type ("ResuRalph is thinking...")
   • Publishes command to SQS queue
   • Discord interaction completes successfully

2. Asynchronous Processing

   • Command Processor Lambda triggered by SQS message
   • Processes long-running operations (AI analysis, PDF diff processing)
   • Uses Discord's follow-up webhook to send results
   • Independent timeout and retry handling

3. Error Handling & Reliability

   • SQS provides built-in retry mechanisms
   • Dead letter queues for failed messages
   • Separate resource pools prevent blocking
   • Cost-optimized: only pay for actual processing time

This pattern is recommended by AWS for serverless asynchronous workflows, using SQS as the message broker between Lambda functions.