I’m Muhammad Saim Ayan, a junior AI engineer and computer science student. I work across the machine learning process: making raw data usable, building and evaluating models, and shaping the results into practical applications.
PROFILE / 01
MUHAMMAD SAIM AYAN
A real-time AI application for image analysis and personalized insights.
I developed the computer-vision workflow around face detection, image preprocessing, and feature extraction, then connected the experience to a secure admin dashboard for verification, user management, and automated reporting.
FaceWell brings image analysis and practical user-management tools together in one application.
Prepare incoming images through preprocessing so they can move into the analysis workflow.
Apply face detection and feature extraction as the foundation for personalized insights.
Support user verification and management with automated reporting for administrators.
Upload a chest X-ray and classify it as Normal or Pneumonia.
Built with an EfficientNet-B0 image-classification model using PyTorch and Torchvision. The end-to-end workflow covers image preprocessing, model training and evaluation, predictions on uploaded images, a FastAPI prediction API, and a Streamlit demo.
A trained model, a prediction API, and an upload interface make the classifier easy to explore.
Train and save a PyTorch model for two chest X-ray classes.
Return class probabilities and mark predictions below 50% confidence as Unknown.
Try an image through the Streamlit interface or send it to the FastAPI endpoint.
An educational portfolio project; not a substitute for medical diagnosis or clinical decision-making.
See my other experiments and code on GitHub ↗
I’m Muhammad Saim Ayan, a junior AI engineer based in Burewala, Pakistan. I’m drawn to the part of AI where careful analysis meets useful outcomes: understanding a dataset, choosing a sensible approach, and learning from how a model performs.
My practice includes data cleaning and visualization, SQL analysis, predictive modeling, image classification, and building with PyTorch and Scikit-learn. I care about writing clean code, communicating findings clearly, and working well with a team.
I’m pursuing a BS in Computer Science at COMSATS University Islamabad, Sahiwal Campus, with expected completion in 2026.
Let’s talk ↗A practical, measured approach
to building with AI.
Clean, explore, and visualize the dataset before deciding what the problem needs.
Prepare useful features, train a model, and choose evaluation methods that fit the task.
Review model performance, tune parameters, and keep experiments organized with tracking tools.
Turn the result into a clear application or report, with maintainable code and useful communication.
Data analysis with Python and Pandas · SQL queries for extraction and aggregation · Version control with Git and GitHub · Team coding challenges and hackathons.
PRESENTTools and techniques I use
to explore, train, and build.
Writing analysis, modeling, and database queries.
Finding patterns and preparing reliable inputs.
Building, measuring, and improving predictive models.
Applying deep learning methods to image-focused tasks.
Tracking experiments and working with model APIs.
Building reproducible projects and collaborating on code.