Faisal Mehmood

Computer Vision Engineer — Visual SLAM & Feature Matching System

Aug 2026 - Sep 2026 • 1 mo

Project: Visual SLAM and Computer Vision Performance Evaluation Developed a Python-based computer vision experimental workflow for evaluating visual SLAM, feature matching, trajectory estimation, and tracking performance. Worked with ORB feature extraction and RANSAC-based geometric filtering to investigate image feature matching under different parameter settings. Evaluated how RANSAC thresholds and ORB feature counts affect tracking performance and computational efficiency. Analyzed trajectory estimation error, feature matching quality, inlier counts, and tracking success rates. Developed reproducible experiments and generated visualizations showing trajectory error versus RANSAC threshold, tracking success rate, estimated FPS, and accuracy-speed trade-offs. Compared original research results with reproduced experimental outputs and documented the observed differences. Organized experimental results into CSV datasets, comparison tables, performance graphs, and technical reports. Technologies: Python, OpenCV, Computer Vision, Visual SLAM, ORB, RANSAC, Feature Matching, Image Processing, Performance Evaluation, Data Visualization.

AI Backend Engineer — Image Processing & Cloud Storage System

Jun 2026 - Jul 2026 • 1 mo

Project: AI Image Processing Backend with FastAPI, PyTorch and Amazon S3 Developed a Python-based backend for an AI image-processing application, combining deep learning model integration with cloud file storage and API services. Integrated a trained UNet++ image segmentation model into the backend processing workflow. Worked with PyTorch model loading and inference-related application components. Developed FastAPI endpoints for application integration, file handling, and backend operations. Implemented Amazon S3 file upload functionality using boto3 and user-specific storage paths. Designed user-based file organization using S3 object prefixes to separate uploaded content. Worked with Amazon S3 Access Points and IAM permissions for controlled file access. Implemented and investigated presigned URL functionality for accessing files stored in S3. Integrated PostgreSQL database connectivity and prepared the backend for containerized AWS deployment. Technologies: Python, FastAPI, PyTorch, UNet++, Image Segmentation, Amazon S3, boto3, AWS IAM, PostgreSQL, Docker, REST APIs.