Skip to content

AtharvaDeokar21/AMD-Stopify

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

31 Commits
 
 
 
 
 
 

Repository files navigation

Team Stopify

🎓 Horizon : AI-Powered Study Abroad Preparation Platform

An intelligent, end-to-end AI platform that transforms how students prepare for international graduate applications.

From profile building to test prep, SOP refinement to research cold emails — this system acts as a structured AI co-pilot throughout the entire journey.

🌍 Vision

To eliminate randomness, missed deadlines, generic applications, and inefficient preparation in the study abroad ecosystem by building a fully adaptive, data-driven, AI-powered preparation engine.

We aim to answer the most important student question:

"What should I be doing today to be ready for Fall 2027?"

And then actually make it happen.

🚀 Unique Value Proposition (USP)

Unlike traditional platforms that provide static resources, this system:

  • Adapts daily schedules dynamically

  • Recalculates study plans based on improvement

  • Generates fresh IELTS writing prompts on demand

  • Scores writing and speaking using AI

  • Detects SOP weaknesses without rewriting the student’s voice

  • Drafts structured, program-specific LORs

  • Generates highly targeted research cold emails

  • Provides actionable resource recommendations

  • Maintains a structured, unified student profile across all modules

This is not just a toolkit. It is a fully integrated AI workflow engine.

🖥️ Usage of AMD Products & Solutions

Our platform is architected around AMD GPUs (ROCm stack) and AMD EPYC processors to enable scalable, cost-efficient AI-driven infrastructure.

🚀 AI Acceleration Layer (AMD GPUs + ROCm)

We leverage AMD GPU acceleration for:

  • Whisper-based speech evaluation

  • Writing & speaking AI inference

  • Batch exam feedback processing

  • Parallel LLM workloads

ROCm-compatible execution ensures low latency, high throughput, and efficient scaling as student usage increases.

🧠 Compute Layer (AMD EPYC)

Our backend runs on AMD EPYC-powered infrastructure to support:

  • Multi-core parallel schedule computation

  • Concurrent diagnostic evaluations

  • Adaptive workload orchestration

  • High-throughput AI services

EPYC’s high core density enables efficient parallel task execution across thousands of users.

⚙️ Scalable & Cost-Efficient by Design

By aligning with AMD architecture, we achieve:

  • Optimized price-to-performance

  • Reduced cost-per-inference

  • Energy-efficient AI scaling

  • Affordable access for students

This infrastructure ensures performance, scalability, and sustainability as the platform grows.

🧠 Core Modules

1️⃣ Profile Intake System (LLM-Powered Conversational Onboarding)

  • Conversational profile collection

  • Intelligent extraction of:

    • CGPA

    • Institution

    • Target country

    • Degree level

    • Intake year

    • Family income

    • Career goal

  • Automatic normalization and validation

  • Follow-up question generation for missing fields

  • Profile completion scoring

Outcome: A unified student profile that drives every downstream module.

2️⃣ Personalized Daily Schedule + Deadline Notifier

Features

  • Works backward from intake year

  • Generates day-level calendar

  • Phase-based planning:

    • Exam phase

    • SOP phase

    • LOR phase

    • Application phase

  • Automatic 2-week buffer before hard deadlines

  • Intensity scaling based on days remaining

  • Urgency-based priority tagging (Green / Yellow / Red)

  • Adaptive rebalancing when tasks are missed

  • Weekly AI digest ("This week matters because...")

Outcome: Students always know what to do today and why it matters.

3️⃣ Test Prep Guide (Adaptive IELTS Engine)

Diagnostic Mini-Test System

  • AI-generated reading passage with MCQs

  • AI-generated writing prompt

  • AI-generated speaking cue card

  • Reading auto-graded

  • Writing scored via rubric-based AI evaluation

  • Speaking analyzed using:

    • Whisper transcription

    • Acoustic metrics (pause, pitch, filler density)

    • LLM structured feedback

Writing Engine

  • Dynamic IELTS-style prompt generation

  • Band scoring across:

    • Task response

    • Coherence

    • Lexical resource

    • Grammar

  • Line-level feedback

  • Generic phrase detection

  • Skill profile auto-update

Speaking Engine

  • Acoustic analysis:

    • Speech rate

    • Long pauses

    • Pitch variation

    • Vocabulary diversity

  • Band estimation heuristic

  • AI coaching feedback

Adaptive Skill Weighting

  • Skill gap calculation

  • Dynamic time allocation per skill

  • Schedule auto-recalibration

Outcome: Structured, measurable progress toward target score.

4️⃣ Resource Recommendation Engine

  • Gap-aware recommendations

  • Weakest skill detection

  • Structured outputs:

    • YouTube resources

    • Practice drills

    • Strategic advice

  • Personalized study guidance

Outcome: Students don’t just know their weaknesses — they know how to fix them.

5️⃣ SOP Assistant (Augmentation, Not Replacement)

We do NOT generate full SOPs.

Instead, we:

  • Score draft using admissions-style rubric

  • Detect generic phrases

  • Evaluate:

    • Career clarity

    • Narrative flow

    • Program fit

    • Writing quality

    • Authenticity

  • Provide line-level suggestions

  • Highlight missing elements

Outcome: Stronger, authentic, program-aligned SOPs.

6️⃣ LOR Assistant

Structured LOR drafting tool.

Input:

  • Recommender type

  • Relationship duration

  • Student strengths

  • Projects

  • Target program

Output:

  • 4–5 paragraph structured LOR draft

  • Specific examples highlighted

  • Professional tone

  • Customizable template

Outcome: High-quality, program-specific recommendation drafts.

7️⃣ Cold Email Assistant (Research Profile Building)

Helps students secure research internships.

Features:

  • Research abstract analysis

  • Student–professor overlap detection

  • Specific technical alignment extraction

  • 4-paragraph structured cold email generation

  • Generic phrase filtering

  • Response probability estimation

Outcome: Highly targeted emails instead of mass mailing.

8️⃣ University & Scholarship Matching Engines

🎯 University Matching Engine

  • Admit probability score

  • Profile improvement simulator

  • Safe / Target / Ambitious classification

  • Faculty alignment NLP scoring

💰 Scholarship Matching Engine

  • Eligibility-based filtering

  • Competitiveness scoring

  • Deadline prioritization

  • Financial planning integration

🏗️ Technical Architecture

  • Backend: Flask

  • Database: PostgreSQL

  • ORM: SQLAlchemy

  • Authentication: JWT

  • LLM Integration:

    • Gemini 2.5 Flash

    • Groq + LLaMA 3

  • Speech Processing:

    • Faster-Whisper

    • Librosa

    • Parselmouth

🔄 Adaptive Intelligence Flow

Profile → Diagnostic → Skill Gap → Schedule → Practice → Re-score → Rebalance → Recommend → Improve

Everything feeds back into the system.

📊 Data Flow Overview

  1. Profile created

  2. Diagnostic establishes baseline

  3. Skill profile updated

  4. Schedule generated

  5. Writing / Speaking practice updates skills

  6. Schedule recalibrates

  7. Resource recommendations adjust

  8. SOP and LOR modules assist application readiness

  9. Cold email supports research profile building

  10. University and scholarship matching guides final applications

🛡️ Ethical Design

  • SOP not auto-written

  • LOR labeled as draft

  • AI band scores clearly marked as estimated

  • Student retains authorship and control

  • No fabricated research claims

📌 What Makes This Different

Most platforms provide:

  • Static test questions

  • Generic SOP templates

  • Non-adaptive schedules

This system provides:

  • Live-generated test content

  • Adaptive scheduling

  • AI-driven skill recalibration

  • Structured feedback loops

  • Unified preparation ecosystem

🎯 Impact

  • Reduces retest costs

  • Prevents missed deadlines

  • Improves application quality

  • Increases research internship success rate

  • Provides clarity and structure in a chaotic process

🧪 Future Enhancements

  • Real-time band projection engine

  • Visa timeline integration

  • Google Calendar sync

  • Scholarship competitiveness scoring

👨‍💻 Built For

  • Undergraduate students targeting MS / PhD

  • MBA aspirants

  • Research-focused applicants

  • Students from emerging education systems navigating global applications

🏁 Conclusion

This project is not a collection of AI tools.

It is an integrated, adaptive preparation system designed to guide students from confusion to clarity — one intelligent decision at a time.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors