PearlAi

The Ultimate Academic Assistant for NPU International Students

AI-powered assistance
24/7 availability
NPU-specific support

Core Features

Designed with precision for the NPU community

AI-Powered Assistance

Specialized in NPU academics, visa guidance, and campus life support with Qwen-2.5-32b model for human-like understanding.

RAG Technology

Combines AI with real university documents using FAISS for lightning-fast retrieval and MiniLM-L6-v2 embeddings for semantic understanding.

Real-time Streaming

Experience fluid, sentence-by-sentence response generation with optimized threading for smooth interaction even on high-latency connections.

Secure & Efficient

Advanced rate-limiting and conversation tracking with automatic API key rotation and robust error handling for 24/7 reliability.

Our Development Journey

From concept to creation: the story behind PearlAi

Creating PearlAi wasn't just a technical challenge—it was a labor of love born from our own struggles as international students at NPU.

Sept 2023

The Inspiration Phase

As international students, we faced daily challenges navigating NPU's systems. Information was scattered across multiple platforms, often only in Chinese. The frustration of seeing fellow students struggle with the same issues sparked an idea: what if we could create an AI assistant specifically for NPU?

  • Gathered feedback from fellow international students
  • Documented common challenges and pain points
  • Researched existing solutions and their limitations
Initial pearl model experiments
July-Aug 2024

Initial Model Experiments

Our journey began with testing different language models, each presenting unique challenges:

  • First attempt: Mistral 7B - Promising but resource-intensive
  • Second try: DeepSeek - Good performance but not optimal
  • Final choice: Qwen - Perfect balance of capabilities and efficiency
  • Many late nights debugging and optimizing prompts
Initial pearl model experiments
September-November 2024

The Development Challenge

Working with limited resources, we faced significant technical challenges that required creative solutions:

  • Utilized Google Colab for model training
  • Dealt with frequent runtime disconnections
  • Struggled with long training times and session limits
  • Developed workarounds for memory constraints
Pearl development process
December 2024-January 2025

The Breakthrough

After months of trials, we finally found our solution:

  • Discovered Qwen model - perfect balance of performance and efficiency
  • Successfully implemented RAG system
  • Overcame context length limitations
  • Achieved significant improvements in response quality
Breakthrough moment
February 2025

Fine-tuning & Optimization

The final stretch involved polishing and perfecting:

  • Refined prompt engineering for better accuracy
  • Improved response time and reliability
  • Enhanced context understanding for NPU-specific queries
  • Countless hours of testing and adjusting
Fine-tuning phase Fine-tuning session in progress
March 2025

Testing Phase

Currently in testing phase with promising results:

  • Positive feedback from initial student testers
  • Continuous improvements and refinements
  • We welcome feedback to make PearlAi even better
Testing phase Testing session with students

Technical Architecture

The engineering behind the experience

User
Frontend
Flask Backend
FAISS Vector DB
Qwen-2.5-32b LLM
Response Generation

Technology Stack

Python
Flask
Langchain
FAISS
HuggingFace
Groq API
JavaScript
HTML5/CSS3

Performance Metrics

1.8s
Average Response Time
95%
Accuracy Rate
3000+
Concurrent Users
99.9%
Uptime

Try PearlAi

Experience the actual PearlAi assistant

Visit PearlAi Website

Infrastructure

The technology powering PearlAi

Server Specifications

Computing Power

2 vCPUs, 4 GiB RAM

Instance Type

ecs.e-c1m2.large

Storage

40 GiB ESSD

Bandwidth

100 Mbps (Pay-By-Traffic)

Operating System

Alibaba Cloud Linux 3.2104 LTS (64-bit)

Why This Configuration?

Optimized for AI API Integration – Handles AI workloads efficiently with Qwen API & Groq

Cost-Effective – Significant savings with long-term subscription plans

Scalable – Suitable for moderate AI traffic, with upgrade flexibility

Reliable Performance – Stable cloud infrastructure for AI applications

Meet the Creators

The minds behind PearlAi

Hana Kouiriti

Hana Kouiriti (@blackt)

Every limitation became an opportunity to innovate.

Yasmine Ait Haddou

Yasmine Ait Haddou (@Hinoko77)

We built the tool we wished we had when we first arrived.

Our Story

We met during our first semester at NPU, both struggling with the same challenges of navigating a new academic system in a foreign country. What began as a mutual frustration evolved into a shared vision: to create a tool that would make life easier for every international student who followed in our footsteps.

Nights of coding turned into weeks, and weeks into months. We balanced our studies with development, often working until dawn to solve particularly challenging problems. When faced with computational limitations, we utilized free resources like Google Colab and optimized our code to work efficiently with the tools available to us.

Today, PearlAi represents not just a technical achievement, but a testament to what determination and friendship can accomplish. We're proud to share it with the NPU community and continue improving it based on your feedback.

The Road Ahead

Our vision for PearlAi's future

Mobile Application

Native apps for iOS and Android to provide on-the-go assistance for NPU students

Q3 2025

Voice Interface

Speak directly to PearlAi for hands-free assistance in multiple languages

Q4 2025

Integrating Maps

integrating smart maps that will guide students around campus and important places in Xi'an.

Q1 2025

NPU Integration

Direct integration with NPU's official systems for personalized schedules and notifications

Q2 2025