ML Engineering Intern

Document Capture: Improve document detection and image alignment for ID cards and passport pages captured using mobile devices. Identify common detection issues and develop solutions using YOLO, image segmentation and OpenCV. Document OCR: Evaluate and improve the accuracy of text recognition when extracting information from MyKad and passports using image preprocessing and OpenCV. Document Fraud: Develop and improve image-based techniques to detect fake identity documents, such as screen replays and photocopies, using colour and texture analysis with NumPy and SciPy. Facial Liveness & Match: Test and improve facial liveness detection and face matching across different devices, lighting conditions and skin tones. Analyse performance and report accuracy and error rates using MediaPipe and face recognition evaluation techniques. On-Device Prechecks: Test and compare lightweight AI models to measure their speed and accuracy on both low-end and high-end mobile devices using TensorFlow.js and MediaPipe. Datasets & Tooling: Organise real-world production failures into labelled test datasets and develop Python scripts to test and compare different AI model versions using pandas and scikit-learn. Requirements: Good Python, and working knowledge of PyTorch and OpenCV. Understanding of CNN and transformer basics, and of how a model is trained and evaluated. Basic statistics: precision and recall, ROC curves, and why a test set must be kept separate. At least one computer vision project (coursework or personal) that you can explain in detail. Clear written English. Nice to have: Image processing or signal processing coursework (colour spaces, Fourier analysis). Experience with detection, segmentation, OCR or face recognition. Docker, Linux, or experience running models on a GPU. Interest in security, fraud or document forensics