ParaPersona by ARCHITYPE

• 

  Landing

• 

  Frame 1

• 

  Frame 2

• 

  Frame 3

• 

  Frame 4

• 

  Frame 5

• 

  Frame 6

ParaPersona is an AI agent as it autonomously understands, reasons, and acts across resume parsing, behavioral analysis, and decision support.

🌟 Inspiration

We were frustrated with the traditional hiring process where candidates are reduced to buzzwords like “hardworking” and “team player,” and recruiters waste time guessing who actually fits the role. We wanted to flip the script by using AI to reveal how candidates think, not just what they claim. That idea sparked ParaPersona: a smart, dual-access platform that delivers clarity on both sides.

🛠️ What it does

ParaPersona is a next-gen AI screening platform that:

• Extracts authentic experience and project work from resumes using Gemini
  
• Creates custom situational questions based on actual resume content
  
• Conducts a 5-question behavioral assessment with tracked response time
  
• Uses timing + text to label responses as instinctive or hesitant
  
• Generates a recruiter-facing report with emotional traits, strengths, and insights
  
• Separates views for Candidates and Recruiters for privacy and flow
  
• Powers everything with a clean, animated UI and a RAG-enabled chatbot

🧩 How we built it

• Frontend: HTML, TailwindCSS, Chart.js (UI + reports + starfield animations)
  
• Backend: Node.js using Gemini 1.5 Pro API
  
• Resume Parsing: pdf-parse to extract raw text
  
• Prompt Engineering: Custom prompts to extract JSON and generate questions
  
• Role Separation: Two-mode system for Test Takers and Recruiters
  
• Final Dashboard: Combined insights + bar and pie charts + AI Q&A interface
  

🐛 Challenges we ran into

• Handling messy or unstructured resume text during parsing
  
• Cleaning Gemini's output to ensure valid JSON structure
  
• Implementing response-time logic in-browser for behavioral scoring
  
• Designing a dual-user interface while keeping the experience seamless
  
• Visualizing emotional insights without oversimplifying user behavior
• Building the frontend exactly as our prototype using HTML

🏆 Accomplishments that we're proud of

• Built a multi-agent Gemini backend with parsing, questioning, and analysis
  
• Clean separation of roles with different access views
  
• Created a scrollable, animated chat interface for the AI
  
• Added bar and pie charts using Chart.js for recruiter-friendly insights
  
• Integrated a dynamic final report showing traits like loyalty, EQ, and results-focus

📚 What we learned

• How to guide LLMs for structured, reliable output using prompt tuning
  
• How response timing can reveal thinking patterns
  
• That clarity and empathy can co-exist in hiring tools
  
• How much polish and UI cohesion impact user trust

🚀 What's next for ParaPersona

• Try making and using our own model instead of Gemini Api
• Instead of using Local Storage, well use mongoDB
• Testing with real users for recruiter + candidate feedback
  
• Packaging it as a recruiting module for future integrations

Built With

• css3
• gemini
• html5
• javascript
• nltk
• rag
• tailwind
• vscode