Saher Sajid
Verified Expert in Engineering
Machine Learning Engineer and Software Developer
Montreal, QC, Canada
Toptal member since July 18, 2023
Saher is an experienced computer vision engineer with an extensive ability to synergize traditional machine learning methods and contemporary approaches. Excelling in object detection, recognition, and visual data synthesis, she has an impressive track record of thriving solutions: creating an innovative speed estimator using stereo vision at Hazen.ai and contributing to Saudi Arabia's e-governance program at Addo.ai. Saher aspires to improve model accuracy and reduce training time continuously.
Portfolio
Experience
- Python - 5 years
- Deep Learning - 4 years
- Machine Learning - 4 years
- OpenCV - 3 years
- PyTorch - 3 years
- Unity3D - 2 years
- C++ - 2 years
- C# - 1 year
Work Experience
Senior Machine Learning Engineer
Hazen.ai
- Developed a vehicle instantaneous speed detection system estimating speed within 10% of the ground truth in the field. The proof of concept employed existing deep neural networks for feature finding and matching to estimate disparity.
- Fine‑tuned a pruned version of SqueezeDet, a fully convolutional deep neural network for real-time object detection for custom classes.
- Trained autoencoder-based neural networks for Optical Character Recognition, Automatic License Plate Recognition (OCR/ALPR). The final model was deployed on edge devices for real-time detection with accuracy upwards of 95% on clean plates.
- Built a synthetic data generation pipeline using Unity 3D's Perception toolkit to provide annotated data for training models such as pose estimation and object detection.
- Ported OpenCV's pattern detection function to Python to develop a camera calibration process that could work within defined constraints.
Machine Learning Engineer
Addo.ai
- Consulted on Saudi Arabia's e-governance program and proposed AI use cases addressing clients' pain points identified during discovery sessions.
- Provided consultation to a travel company to improve lead conversion rates. The project focused on developing a strategy enabling call center agents to prioritize their efforts on leads more likely to convert into customers.
- Designed the migration of a revenue assurance platform for Telenor, a telecom company. This was part of the data platform solution for Telenor's switch to big data platforms from legacy DBMS systems like Teradata.
Project History
Vehicle Instantaneous Speed Using Stereo Vision
The project's initial phase involved calibrating the stereo system, which I accomplished using OpenCV. To ensure the accuracy of the calibration parameters, I utilized the MATLAB calibration tool as a secondary verification method.
The core of the system's functionality lies in its use of deep neural networks for feature finding and matching. These networks implemented using PyTorch were integral in estimating depth, a critical factor in determining vehicle speed.
Education
Master's Degree in Artificial Intelligence
Queen Mary, University of London - London, United Kingdom
Bachelor's Degree in Computer Engineering
College of Electrical and Mechanical Engineering, NUST - Rawalpindi, Pakistan
Skills
Libraries/APIs
PyTorch, OpenCV
Languages
Python, C++, C#, Embedded C++
Frameworks
Unity3D
Platforms
Linux, Docker
Other
Deep Learning, Machine Learning, Computer Engineering, Artificial Intelligence (AI), Game AI, Natural Language Processing (NLP), Embedded Systems
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
Share your needs
Choose your talent
Start your risk-free talent trial
Top talent is in high demand.
Start hiring