GradCompass - AI-Powered Graduate Application Assistant

Multi-agent LLM system for personalized graduate school guidance and application support

Quick Navigation: Overview • System Architecture • Technical Components • Key Features • User Experience • Impact & Use Cases • Future Enhancements

Overview

Built a comprehensive multi-agent LLM system delivering personalized graduate school guidance, document evaluation, and recommendations through an 8-agent framework with specialized capabilities.

Status: Completed

Objective

Design and implement an intelligent graduate application assistant that provides:

  • Personalized university recommendations
  • Document evaluation and feedback (SOP, CV, essays)
  • Strategic application planning and advising
  • Interview preparation and simulation
  • Comprehensive guidance throughout the application journey

System Architecture

8-Agent Framework

Each agent has a specialized role with dedicated LLM configuration:

  1. University Recommendation Agent
    • Matches applicant profiles to suitable programs
    • Considers academic background, research interests, career goals
    • Factors in location preferences, funding, and program strengths
  2. Statement of Purpose (SOP) Evaluator
    • Analyzes narrative structure and coherence
    • Assesses research interest articulation
    • Provides detailed improvement suggestions
    • Checks alignment with program requirements
  3. CV/Resume Reviewer
    • Evaluates experience presentation
    • Suggests formatting improvements
    • Identifies missing or underemphasized elements
    • Tailors advice to academic standards
  4. Essay Feedback Agent
    • Reviews supplemental essays and personal statements
    • Evaluates authenticity and voice
    • Suggests content strengthening strategies
    • Checks for common pitfalls
  5. Strategic Advising Agent
    • Develops application timeline strategies
    • Recommends program mix (reach/target/safety)
    • Advises on application prioritization
    • Provides funding and fellowship guidance
  6. Requirements Checker
    • Tracks program-specific requirements
    • Monitors application deadlines
    • Ensures completeness of materials
    • Flags missing components
  7. Interview Preparation Agent
    • Generates common interview questions
    • Provides question-answering strategies
    • Offers feedback on practice responses
    • Simulates interview scenarios
  8. Interview Simulator
    • AI-powered mock interview conductor
    • Adaptive questioning based on responses
    • Real-time feedback on answers
    • Behavioral and technical question coverage

Technical Components

RAG (Retrieval-Augmented Generation) Pipeline

Knowledge Base:

  • University program information and requirements
  • Admission statistics and trends
  • Research group descriptions and faculty profiles
  • Application guidelines and best practices
  • Sample successful application materials

Retrieval System:

  • Semantic search over program databases
  • Context-aware document retrieval
  • Dynamic knowledge injection into agent prompts
  • Real-time information updates

Integration:

  • Each agent accesses relevant knowledge subsets
  • Personalized retrieval based on user profile
  • Reduces hallucination through grounded responses
  • Maintains factual accuracy in recommendations

Multi-Agent Orchestration

Agent Communication:

  • Shared context across agents for consistency
  • Information passing between specialized agents
  • Coordinated workflow for complex queries
  • User session management

Role-Specific LLMs:

  • Different prompting strategies per agent
  • Specialized system prompts for each role
  • Temperature and parameter tuning per agent
  • Output format standardization

Key Features

Personalized Recommendations

University Matching:

  • Profile-based program suggestions
  • Research fit assessment
  • Career goal alignment
  • Financial aid considerations

Strategic Planning:

  • Custom application timelines
  • Program portfolio optimization
  • Resource allocation guidance
  • Deadline management

Document Evaluation

Comprehensive Feedback:

  • Structural analysis of application materials
  • Content quality assessment
  • Specific improvement recommendations
  • Iterative refinement support

Best Practice Guidance:

  • Field-specific conventions
  • Program-specific tailoring
  • Common mistake avoidance
  • Successful example patterns

Interview Preparation

AI-Powered Simulation:

  • Realistic interview environment
  • Adaptive question selection
  • Performance feedback
  • Confidence building exercises

Question Bank:

  • Technical questions by field
  • Behavioral competency questions
  • Research discussion scenarios
  • Program-specific inquiries

User Experience

Workflow

  1. Profile Creation: User inputs background, interests, goals
  2. Initial Consultation: Strategic advising agent provides overview plan
  3. Program Discovery: Recommendation agent suggests target programs
  4. Document Drafting: User develops application materials with feedback
  5. Iterative Refinement: Document evaluator agents provide multiple rounds of feedback
  6. Interview Prep: Practice with simulator agent before real interviews
  7. Application Monitoring: Requirements checker tracks progress and deadlines

Interaction Modes

  • Conversational Interface: Natural language queries and responses
  • Document Upload: Direct material submission for evaluation
  • Guided Workflows: Step-by-step application process support
  • On-Demand Advice: Quick answers to specific questions

Technical Implementation

Backend:

  • Multi-agent orchestration framework
  • RAG pipeline with vector database
  • LLM API integration (GPT-4, Claude, etc.)
  • Session and context management

Frontend:

  • User-friendly chat interface
  • Document upload and management
  • Progress tracking dashboard
  • Recommendation visualization

Data Management:

  • User profile storage
  • Application material versioning
  • University database maintenance
  • Feedback history tracking

Impact and Use Cases

Target Users

  • Undergraduate students planning graduate applications
  • International students navigating US/UK application systems
  • Career changers pursuing advanced degrees
  • Researchers seeking PhD programs

Value Proposition

Time Savings:

  • Automated program research and filtering
  • Quick document feedback cycles
  • Centralized application tracking

Quality Improvement:

  • Expert-level advice at scale
  • Multiple perspective document reviews
  • Data-driven recommendations

Accessibility:

  • 24/7 availability
  • Cost-effective compared to consultants
  • Reduces information asymmetry

Future Enhancements

Feature Additions:

  • Integration with university application portals
  • Automated letter of recommendation requests
  • Post-admission decision support (offer comparison)
  • Alumni network connections

Technical Improvements:

  • Fine-tuned models for specific agent roles
  • Expanded knowledge base coverage
  • Multi-modal support (video interview simulation)
  • Personalized learning from user feedback

Ethical Considerations:

  • Transparency about AI limitations
  • Human expert review integration
  • Bias mitigation in recommendations
  • Privacy protection for user data

Conclusion

GradCompass demonstrates the potential of multi-agent LLM systems to provide comprehensive, personalized support for complex, multi-stage processes like graduate school applications. By combining specialized agents with retrieval-augmented generation and intelligent orchestration, the system delivers value across the entire application journey—from initial planning through interview preparation.

This project showcases how AI can democratize access to high-quality educational advising, helping students navigate complex application processes with confidence.