I design, train, and deploy Machine Learning models — from Computer Vision and NLP to predictive analytics — to solve real business problems.
Outcome-focused AI/ML solutions. I translate business problems into deployable models.
Detection, classification, and tracking for real-world images/videos (YOLO, CNNs, transfer learning).
Text classification, summarization, and retrieval-augmented workflows tailored to domain data.
Supervised/unsupervised ML for risk scoring, forecasting, and decision support.
Selected work. Each card links to code, write-ups, or a filtered view on GitHub.
Binary classifier to detect misinformation; metrics, pipeline, and dataset handling included.
CNN-based facial emotion recognition with training curves and evaluation dashboard.
Image classifier for common crop diseases; transfer learning and augmentation.
GTSRB dataset with strong accuracy and robust preprocessing.
Classical + ML approach to robust lane detection on dashcam footage.
Deployed model to classify masked vs unmasked faces in real-time.
Short, business-focused breakdowns. Replace with real client work as you go.
Problem: Manual verification was slow and inconsistent.
Solution: Trained a TF‑IDF + linear model pipeline with calibrated thresholds.
Result: ~92% accuracy on validation; reduced manual checks by ~60% in tests.
Problem: Need for accurate sign detection in embedded environments.
Solution: Efficient CNN with data augmentation and quantization-aware training.
Result: High accuracy on GTSRB and smooth inference on CPU.
I’m an AI/ML engineer from Karachi, focused on building practical models for Computer Vision, NLP, and classical ML. I love clean experiments, reproducible code, and translating business needs into measurable metrics.
I share AI news, tutorials, and project breakdowns.
Have an AI/ML problem to solve? Let’s talk.
/assets/Muhammad_Rayan_Shahid_Resume.pdf and case study template to /assets/case-study-template.pdf. Replace avatar and OG image in /assets/.