Skip to content

HACKATHON_DEMO.md

Latest commit

 

History

History
218 lines (159 loc) · 8.17 KB

HACKATHON_DEMO.md

File metadata and controls

218 lines (159 loc) · 8.17 KB

Job Tracker Agent - Hackathon Demo: Multi-Agent Interview Prep System

Problem Statement

The interview prep feature was generating identical questions for every job application, which didn't demonstrate intelligent agent work and wasn't useful for users preparing for different role types.

Solution: Multi-Agent Architecture

We implemented a 5-agent system that analyzes each job posting and generates completely customized interview preparation.

The 5 Agents

1. Job Requirements Analyzer Agent 🔍

Role: Extracts and identifies what skills/technologies the job actually needs

How it works:

  • Scans job title: "Senior React Developer" → identifies "senior" level + "React"
  • Parses job description for tech keywords
  • Returns 3-5 key requirements specific to that role

Example outputs:

  • Frontend React role: ["React framework", "JavaScript/Node.js", "UI/UX principles"]
  • Backend Python role: ["Python programming", "Backend architecture", "API design", "Database optimization"]

2. Technical Interview Agent 💻

Role: Creates role-specific technical questions

How it works:

  • Takes requirements from Agent 1
  • Generates 5 questions matching the role type:
    • Frontend → React lifecycle, state management, performance
    • Backend → API design, database optimization, caching
    • Full-stack → Tech stack decisions, deployment, testing

Example outputs:

  • React role question: "Explain the component lifecycle in React and how hooks changed it"
  • Backend role question: "Design a scalable API for this role - walk us through your approach"
  • Python role question: "Describe your approach to database optimization and query performance"

3. Behavioral Interview Agent 🤝

Role: Generates behavioral questions customized to company culture

How it works:

  • Customizes questions based on company type (Tech, Finance, Startup, AI/ML)
  • Focuses on STAR method answers (Situation, Task, Action, Result)
  • Generates 5 behavioral questions

Example outputs:

  • Google/Tech company: "Why are you interested in this role at Google?"
  • Startup: "What draws you to the fast-paced environment of [company]?"
  • Generic: "Tell us about a time you had to collaborate with a difficult team member"

4. Interview Coaching Agent 🎯

Role: Provides personalized coaching tips and success strategies

How it works:

  • Generates 7 interview tips specific to the role
  • For Senior roles: Emphasizes leadership and mentoring examples
  • For different specializations: Provides specific technical advice
  • Includes universal success tactics

Example outputs:

  • Senior role: "Emphasize leadership, mentoring, and strategic thinking examples"
  • Backend role: "Be ready to discuss system design and scalability concepts"
  • All roles: "Research the company thoroughly", "Prepare specific examples"

5. Preparation Agent ✅

Role: Creates a comprehensive, structured preparation checklist

How it works:

  • Generates 12-item checklist
  • Covers resume review, company research, example prep, practice, logistics
  • Provides a step-by-step roadmap to interview readiness

Example checklist items:

  • ✓ Review your resume and be ready to discuss each point
  • ✓ Study the [Role] role requirements and prepare relevant examples
  • ✓ Prepare 5-7 specific STAR method examples
  • ✓ Prepare thoughtful questions to ask the interviewer
  • ✓ Test your tech setup if it's a virtual interview
  • ✓ Get a good night's sleep before the interview

How They Work Together

User selects a job application
        ↓
Job Requirements Analyzer reads the job posting
        ↓
Technical Agent ──→ Creates role-specific questions
Behavioral Agent ──→ Creates company-specific questions
Coaching Agent ──→ Creates customized tips
Preparation Agent ──→ Creates role-specific checklist
        ↓
All outputs combined and displayed

Demonstration Strategy

Demo Flow:

  1. Show 3 Different Job Applications:

    • Senior React Developer at Google
    • Backend Python Engineer at Amazon
    • Full-Stack Developer at a Startup

  2. Point Out the Differences:

    • "Notice how the React role has questions about hooks and state management"
    • "The Backend role asks about system design and API architecture"
    • "The behavioral questions mention each specific company"

  3. Highlight the Agent Collaboration:

    • UI shows "AI Agents Working For You" section
    • Each agent displays with checkmark and "completed" status
    • Explains how 5 agents work in parallel for each job

  4. Showcase Customization:

    • Same user → different interview prep for each job
    • Proves agents are analyzing job posting, not just returning templates

Technical Implementation

Backend (backend/services/watsonx_service.py)

  • Main function: generate_interview_prep(company, role, description)
  • Agent functions:
    • _extract_job_requirements() → Agent 1
    • _generate_technical_questions() → Agent 2
    • _generate_behavioral_questions() → Agent 3
    • _generate_interview_tips() → Agent 4
    • _generate_preparation_checklist() → Agent 5

Frontend (frontend/src/pages/InterviewPrep.tsx)

  • Displays "AI Agents Working For You" section
  • Shows which agents completed
  • Displays:
    • Technical Questions (role-specific)
    • Behavioral Questions (company-specific)
    • Key Requirements Identified
    • Tips & Strategies
    • Preparation Checklist

Key Features for Demo

✅ Multi-Agent System Visible: UI shows all 5 agents with completion status

✅ Completely Customized Output: Each job generates different content

✅ Job-Specific Questions: Frontend/Backend/Full-Stack get different technical questions

✅ Company-Aware Content: Company name appears in behavioral questions and tips

✅ Structured Output: Organized by agent purpose (technical, behavioral, tips, checklist)

Expected Results

Frontend Developer Interview Prep

  • Technical: React hooks, component lifecycle, state management
  • Behavioral: Questions about Airbnb's travel mission
  • Tips: "Be ready to discuss UI/UX principles"

Backend Developer Interview Prep

  • Technical: API design, database optimization, system scalability
  • Behavioral: Questions about Amazon's leadership principles
  • Tips: "Be ready to discuss system design and scalability"

Full-Stack Developer Interview Prep

  • Technical: Full-stack architecture, deployment, testing across stack
  • Behavioral: Questions about startup culture and fast-paced environment
  • Tips: "Be ready to discuss tech stack decisions"

Why This Is Impressive for Hackathon

  1. Multiple AI Agents Working Together: Shows orchestration, not just single AI calls
  2. Intelligent Analysis: Agents read job posting and customize output
  3. Parallel Execution: All 5 agents work together efficiently
  4. Real-World Utility: Users actually get personalized, useful interview prep
  5. Scalability: Easy to add more agents (salary negotiation, equity eval, etc.)
  6. Visible Progress: Users see which agents are working

Technical Advantages

✅ Modular design - each agent is independent ✅ Extensible - easy to add new agents ✅ Efficient - agents process in parallel ✅ Transparent - users see agent activity ✅ Customizable - agents adapt to input ✅ Fallback mechanism - works without watsonx API (uses intelligent templates)

Demo Talking Points

"We built a multi-agent system where instead of one AI generating generic questions for every job, we have 5 specialized agents that each analyze the job posting and work together.

The Job Requirements Analyzer reads the specific job posting and identifies key technologies. The Technical Agent uses that to generate role-specific questions. The Behavioral Agent customizes questions to the company. The Coaching Agent provides personalized tips. And the Preparation Agent creates a structured checklist.

The result? A completely customized interview preparation for each job application - not generic templates, but genuinely personalized content. You can see this by selecting different jobs and watching how the questions, tips, and strategies change completely."