AI Projects
BiL-FaND
Ensemble-based system detecting fake news in English and Urdu through text, data, and multimedia analysis.
Key Features
- Textual analysis layer using Multilingual BERT achieving 87% accuracy
- Categorical data processing with LSTM achieving 97% accuracy
Project Resources
Asthma Prediction
Early detection of worsening asthma aims to predict daily asthma related ED visits and admissions.
Key Features
- Textual analysis layer using Multilingual BERT achieving 87% accuracy
- Categorical data processing with LSTM achieving 97% accuracy
Project Resources
Smart Shopping
Enables brick and mortar retailers to track customer interests to better compete with online retail.
Key Features
- Provides unprecedented insight into customer behavior
- Helps and enhance optimize sales processes
- Improves in-store shopping experience overall
Project Resources
Forensic Intelligence
Analyzes forensic casework data to identify patterns in forensic science conducted.
Key Features
- Improves crime prevention
- Helps health provSupports Recovery from crimeiders take proactive measures
- Enhances forensic effectiveness
Project Resources
Skin Lesion Classification
Classifies skin lesions using CNNs and feature reduction with 95.8% accuracy.
Key Features
- Handles class imbalance in skin datasets
- Combines DarkNet and Inception V3
- Feature reduction with Moth Flame Optimization
Project Resources
AI Companion
An AI-powered companion that answers questions and engages in conversations.
Key Features
- Engages in natural language conversations
- Provides the best relevant answers
- Learns and adapts on user interactions
Project Resources
Saraiki Language Word Prediction
Predicts next words in Saraiki using Word2Vec and Roberta language models.
Key Features
- Supports regional language NLP
- Evaluates context-based word similarity
- Compares multiple embedding techniques
Project Resources
User Interest Forecasting
Identifies and predicts user interests over time using LDA, BERT, and ARIMA.
Key Features
- Forecasts user interest using social data
- Topic modeling via LDA
- Interest-level prediction using ARIMA
Project Resources
Cancer Detection
Task-Specific Fine-Tuning via Bottleneck for Weakly-Supervised Pathology Whole Slide Image Classification.
Key Features
- Weakly-supervised pathology classification using variational information bottleneck.
- Top-k patch extraction
- End-to-end training pipeline for whole image classification with deep learning.
Project Resources
RAG Document Q&A System
A Retrieval-Augmented Generation (RAG) system for document question answering and summarization.
Key Features
- Ask questions about your documents and get accurate answers with source attribution
- Generate comprehensive summaries of uploaded documents
- Process PDF, DOCX, Excel (XLSX/XLS), and CSV files
Project Resources
VoxAI
An AI-powered platform that converts voice commands into SQL queries with a modern web interface.
Key Features
- Convert spoken commands to text using OpenAI Whisper
- Transform natural language into SQL queries
- Intelligent chat capabilities using local LLM models
Project Resources
VocalCraft-AI
Generates editable promotional pamphlets through voice commands.
Key Features
- Creates editable pamphlets via voice
- Boosts marketing efficiency
- Easy and fast content generation