-
Model 1: Flight Delay Prediction
Flight delays can disrupt passenger schedules, decrease efficiency, increase capital costs, and require reallocation of flight crews and aircraft. This can negatively affect passenger demand.
Purpose: This model aims to predict the estimated duration of flight delays for each flight, helping airlines and passengers plan ahead and reduce the impact of delays.
Key Features: Accurate prediction of delays will allow airlines to implement action plans, reducing time, capital, and resource losses.
Airline: This model is based on flight data from Tunisair, the national airline of Tunisia.
Technologies Used: The model was trained using Ridge regression, a regularization technique that helps reduce overfitting in machine learning models, along with other Machine Learning techniques.
Evaluation: The model’s performance is evaluated using Root Mean Square Error (RMSE), providing accurate delay time estimates.
-
Model 2: Natural Language Processing
Our NLP model powers sentiment analysis tools used in social media monitoring, helping brands understand customer feedback by analyzing text data at scale.
Technologies Used: K-means clustering, Random Forest, Python, Scikit-learn, Pandas
Key Impact: Improved customer engagement by 25%, reduced marketing costs by 15%.
-
Model 3: Predictive Analytics
This model is deployed in retail to forecast demand, helping businesses optimize stock levels and reduce waste while ensuring product availability for customers.
Technologies Used: K-means clustering, Random Forest, Python, Scikit-learn, Pandas
Key Impact: Improved customer engagement by 25%, reduced marketing costs by 15%.
-
Modal 4: Computer Vision Classification
Project Overview: In this hands-on project,
we built an image classification model to identify cassava leaf diseases using a subset of the dataset from Dataverse.
The goal was to classify cassava leaves into three categories: CBSD, CMD, and Healthy.
The dataset was cleaned, prepared, and split into training and testing sets. Simple EDA was performed to understand the data distribution and identify potential anomalies.
Model Training: The model was trained using FastAI and ResNet34 as the backbone.
we built an image classification model to identify cassava leaf diseases using a subset of the dataset from Dataverse.
Model Evaluation: After training, the model was evaluated using metrics like accuracy, precision, and recall.
Confusion matrices and loss visualizations were plotted to better understand the model's performance on different classes.
Technologies Used: FastAI, PyTorch, ResNet34, Data Augmentation.
AI/ML Projects
Completed Projects
-
Project Customer Segmentation
An AI-powered customer segmentation system for a large retail business. It uses machine learning models to classify customers based on their buying behavior, providing personalized marketing insights.
Technologies Used: K-means clustering, Random Forest, Python, Scikit-learn, Pandas
Key Impact: Improved customer engagement by 25%, reduced marketing costs by 15%.
-
Project Maternal Risk
This project aims to predict maternal health risks based on IoT monitoring data collected from various hospitals and maternal health care centers in Bangladesh. The model identifies key health indicators that signal risks during pregnancy.
Technologies Used: XGBoost, Pandas, Seaborn, Scikit-learn, Gradio, Pandas Profiling
Key Features:
- Data attributes such as age, blood pressure (Systolic and Diastolic), blood glucose levels, and heart rate
- Risk level prediction based on health indicators
Key Impact: This model assists in early detection of potential maternal health risks, allowing healthcare providers to take preventive measures and improve pregnancy outcomes.
-
Project Breast Cancer
This project focuses on developing a machine learning model to predict breast cancer based on various diagnostic attributes. The goal is to aid healthcare professionals in making timely and accurate decisions, improving early detection and treatment planning.
Technologies Used: Support Vector Machines (SVM), Random Forest, Logistic Regression, Scikit-learn, Pandas, NumPy, Matplotlib
Key Impact:
- Improved diagnostic accuracy for breast cancer.
- Enabled early detection, reducing the need for invasive procedures.
- Streamlined treatment workflows by providing predictive insights.
Evaluation Metric: Accuracy, F1 Score
Ongoing Projects
-
Project TaskEd
A task management app that allows users to sort projects by priority categories such as "important," "personal," and "work" tasks. The app features a to-do list that can be updated with statuses like "completed" and "pending."
Technologies Used: Flask, Python
Current Progress: Basic task sorting and status updating functionalities have been implemented. Integration with user accounts is finished.
-
Project AlphaHealth
An AI-driven healthcare app that helps users navigate healthcare services. It allows users to find appropriate healthcare centers within their location, view available medical tools, check available medical staff, and rate medical officers. Users can choose and get in touch with medical officers for discussions or consultations. The app keeps track of medical records, enables users to schedule appointments, and offers a search bar for researching medical terms.
Technologies Used: AI, Python (Flask), Google Maps API, Firebase (Firestore), Natural Language Processing (NLP)
Current Progress: Initial UI/UX design completed, basic navigation structure implemented, and integration with Firebase for user data and appointment scheduling is ongoing.
-
Project Gamma
A natural language processing (NLP) chatbot designed to handle complex customer queries in real time for a telecom provider.
Technologies Used: BERT, Dialogflow, Python, Google Cloud
Current Progress: The model development is 15% complete, with initial phases of training and data preprocessing in progress.
-
Project Fraud Detection
A real-time fraud detection system for an online payments platform, designed to identify suspicious transactions and prevent fraud before it occurs.
Technologies Used: Isolation Forest, Logistic Regression, Python, PyTorch
Current Progress: The modal is yet to be started.