Navid Yousefabadi

• Leveraging data and LLMs for creating products in .NET Maui that would change human life's for better.

200 hours University of Calgary internship
1. Image Annotation Efficiency Enhancement:

• Spearheaded an initiative that drastically reduced image annotation workload by approximately 80%.
• Leveraged PyTorch and CUDA to train machine learning models, fine-tuned hyperparameters for optimal performance.
• Deployed the solution seamlessly using Nuclio serverless framework.
• Introduced an automated image annotation process via CVAT integration, transforming the way data was labeled.

2. Critical Obstruction Detection Overhaul:

• Single-handedly resolved a pivotal issue within the stereo camera software by eliminating persistent false positives in the obstruction detection feature.
• Innovatively designed and implemented an advanced obstruction detection algorithm, leading to a remarkable enhancement in accuracy.
• Successfully integrated the long-awaited obstruction alarm feature to the software by making comprehensive backend and frontend adjustments, demonstrating exceptional problem-solving and development skills.

• Spearheaded a transformative initiative at Kuva Systems Inc., developing and implementing sophisticated machine learning pipelines that revolutionized processes and substantially bolstered revenue.
• Leveraged the power of Azure SDK to construct robust pipelines, applying advanced classification techniques to predict gas image patterns and regression models for precise quantification of gas quantities.
• Demonstrated technical prowess by successfully deploying these cutting-edge machine learning models as web endpoints on Azure Kubernetes Service, ensuring seamless integration and real-time availability.

• Successfully improved Pembina's oil & gas pipeline integrity and decreased costs at least 300% by predicting pipeline failures with deep learning algorithms
• Handling imbalanced data (SMOTE, ADASYN, SVC, etc), and multimodal data
• Applying a wide variety of ML techniques to improve performance in various classification and regression problems; data generation (Generative adversarial networks, etc), feature engineering for structured and unstructured data, optimization of deep learning parameters.

• Successfully decreased Electromagnetic Compatibility (EMC) fault detection cost by more than %100 by visualizing the electromagnetic field radiation of devices under test in OpenGL (in C++), and added the functionality to a .NET application written in C#

• Established a framework for statistical hypothesis testing (e.g T-test, bootstrap test, permutation test, etc.) to discriminate between physics theories.

• High performance simulation of adiabatic quantum computers (Object Oriented Programming in C++)