Forough Karandish

Computer Vision & Computer Graphics Research Engineer with 7+ years of experience spanning autonomous driving, 3D machine learning, and applied AI research, from academic research (M.Sc., Concordia University) to industry deployment (Huawei, Vyoo AI). Specializes in deep learning for 3D/2D data, multi-sensor (LiDAR, Radar, Camera) perception, synthetic data generation, and production CV pipelines, with peer-reviewed publications in computer vision for safety applications.

Computer Vision Research Engineer — Autonomous Driving at Huawei Technology Inc. (2021-09 – 2023-09)

• Researched and developed deep learning and computer vision techniques for autonomous driving, including 3D rendering/augmentation and 3D part segmentation; substantially improved 3D object retrieval accuracy by generating synthetic imagery and building a new training dataset.
• Applied Vision Transformer (ViT)-based multi-camera fusion to combine features from multiple camera views into a unified bird's-eye-view scene representation for perception.
• Designed and developed self-driving car pipelines, partnering with planning and perception teams to integrate CV solutions into the broader AD system.
• Built 2D object detection, segmentation, and tracking models; applied self-supervised learning, multi-view CNNs, and embedding spaces to point cloud and polygon mesh data.
• Worked with multisensory input data (LiDAR, Radar, Camera) for autonomous driving perception.
• Fine-tuned models and ran sensitivity analysis to improve robustness for production self-driving pipelines.
• Researched knowledge graph and reasoning-based approaches for autonomous driving applications.

3D Machine Learning Developer at Vyoo AI (2020-11 – 2021-09)

• Researched and built computer vision systems for graph-based and point cloud 3D data, including 3D classification models (DGCNN, PointNet, Multi-view CNN) implemented in TensorFlow and PyTorch, with GPU-accelerated training in Docker.
• Designed and developed a systematic synthetic RGB-D image generator in Blender to produce training data for 3D object classification.
• Built pre- and post-processing pipelines using Open3D, OpenCV, NumPy, and scikit-learn: data augmentation for synthetic training data, and point cloud de-noising/RGB-D registration for evaluating real-world data.
• Evaluated model performance across multiple datasets through sensitivity analysis.
• Maintained code repository on Bitbucket and authored technical documentation covering model architecture, implementation steps, results, and challenges.

Data Scientist (Computer Vision Developer) at Moov AI (2019-04 – 2019-10)

• Built a human body posture recognition system using TensorFlow PoseNet for joint and posture analysis.
• Developed a machine vision banking system using Optical Character Recognition (OCR) to extract text from images for downstream ML analysis.
• Optimized real-time online advertising bid pricing using XGBoost, engineering features from location, time, and demographic data.
• Built a fraud detection model using scikit-learn to flag fraudulent transactions on bank cheques.
• Extracted analytical data features and developed, evaluated, and optimized machine learning and neural network model architectures, benchmarking performance metrics against design feasibility.

Computer Vision Researcher (Intern) at Indus.ai (2018-01 – 2018-05)

• Built a real-time object classification and detection system to identify potential hazards for construction workers.
• Trained, evaluated, and selected features across TensorFlow object detection models (SSD MobileNet, RCNN, ResNet).
• Researched and improved TensorFlow Object Classification API architecture, refining layer design, activation functions, loss functions, regularization, optimization, and computational complexity.

Computer Vision Researcher (Intern) at GreenOwl Mobile (2016-09 – 2017-03)

• Built an object detection and localization model using CNNs — HOG-based in MATLAB and SSD MobileNet in Python.
• Generated a large-scale synthetic RGB image dataset using 3D models in 3ds Max, varying weather conditions and object position/color.
• Researched deep learning, CNNs, and computer graphics techniques, benchmarking CNN architectures to identify the best fit for image classification.

M.Sc. in Systems Engineering (Computer Vision) – Concordia University (2016-09 – 2019-04)

B.Sc. in Electrical Engineering (Telecommunications) – Fars Science and Research University (2010-09 – 2014-05)