Machine Learning Engineer - Computer Vision

About the role

An AI/ML team in Islamabad is looking for a Machine Learning Engineer with 1-2 years of experience building production machine learning systems, with a strong computer vision focus. You will build a computer vision pipeline that automatically extracts structured digital data from scanned images, covering preprocessing, layout and text detection, segmentation, post-processing and data export, with accurate results on real-world, imperfect inputs.

Key responsibilities

• Design, build and deploy an end-to-end computer vision pipeline for extracting structured data from scanned images.
• Train and fine-tune deep learning models for segmentation, detection and classification (e.g., U-Net, SAM-based or instance segmentation models).
• Prepare, annotate, augment and generate image datasets, including synthetic data, for training and evaluation.
• Develop image processing and post-processing methods such as dynamic programming, tracking and graph-based methods.
• Apply supervised, unsupervised and self-supervised learning for feature extraction, quality estimation and continuous improvement.
• Evaluate on hand-verified samples using accuracy metrics beyond IoU and Dice, and optimise for accuracy, performance and latency.
• Write clean, modular, tested code and contribute to deployment, monitoring and model versioning.
• Collaborate with engineers and domain experts on requirements, edge cases and quality checks.

Required profile

• 1-2 years of experience developing and deploying production-level machine learning algorithms.
• Minimum BSc in Computer Science; a Master's in AI/ML, Computer Science, Data Science or a related field is preferred.
• Location: Islamabad (on-site).

Required skills

• Strong Python and software development principles (OOP, testing, Git).
• Training deep learning models in PyTorch and/or TensorFlow, including segmentation or detection models.
• Image processing fundamentals (OpenCV, scikit-image): deskewing, thresholding, morphology and geometric transforms.
• Data preparation, augmentation and evaluation for computer vision models.
• Production ML systems with basic MLOps practices.
• Preferred: OCR tools (PaddleOCR, Tesseract, TrOCR), instance segmentation, synthetic data generation, tiling large images and GPU training.