Sameer Zaheer Email & Phone Number

Toronto, On, Ca

Toronto, Ontario, Canada

Toronto, Canada Area

■ Developed an end-to-end machine learning (ML) pipeline for keyword spotting,comprising of data processing, training neural networks (NNs), model deployment,scaling pipeline to produce 100+ models semi-automatically. Including:--‣ Improved accuracy of DNN, CNN and RNN by 30% by automating NN hyperparameter search--‣Optimized C++ inference code, reducing CPU and memory usage by 40%--‣Containerized model training and testing (Docker) and deployed it on a kubernetes cluster--‣Lead daily standups. Onboarded 5 developers, 2 project managers, mentored 2 non-developers as they learned development.-------------------------------------------■ Improved the accuracy of French speech recognition (ASR) by 5% by modifying thelanguage model to infer liaison pronunciations from ngram context

Toronto, Canada Area

■ Implemented computer vision algorithms and models (C# and C++) for segmentation, classification and labeling of medical images (CT and MR). This included segmentation of lesions, object detection and labeling of vertebrae, object detection of femur head and acetabulum cup, and detection of pelvic landmarks.■ Optimized low-level code (C# and C++) for efficient computation: relating to raytracing, radial transform, graph search, bit manipulation for image masking, etc.■ Maintained the Withinsight Framework (WIF), a versatile SDK (written in .NET) for image processing and visualization. This included debugging, maintaining a usable API and implementing CI/CD and unit tests.

Toronto, Canada Area

■ Developed a computer vision algorithm (C# and C++) for segmentation, classification andlabeling of spinal CT images. ■ Improved time performance of algorithm by eliminating bottlenecks, and parallel computing

Toronto, Canada Area

Designed and implemented (in C++), a computer vision algorithm for segmenting vessels in MR images in near real-time; this worked jointly with registration for image fusion.

Toronto, Canada Area

Taught matrix algebra using a problem-based learning approach with by MATLAB.(ESC103)

Singapore

Developed and tested an algorithm to separate two apparently touching vessels in CT images of the liver. The algorithm was successful (visually inspected) in 10 of the 11 cases tested.

Toronto, Canada Area

Synchronized and tested multiple signal generators with a digitizer to suppress jitter and eliminate trace dropouts. System to be used for research intra-vascular ultrasonography.

Toronto, Canada Area

Designed and implemented a synchronized experiment measuring electroencephalography and electrocardiography while performing mathematical tasks. Successfully found EEG frequency and ERP indicators of task complexity, in tasks with increasing memory load.

Masc

Basc