David Szczecina
Machine Learning Researcher at the UW Vision and Image Processing Lab.
Systems Design Engineering MASc at University of Waterloo.
Mechatronics Engineering BASc at University of Waterloo
Artificial Intelligence Researcher, Vision and Image Processing Group, (November 2023 - Current)
- Published multiple award winning AI research papers at Conference of Vision and Intelligent Systems
- Developed method for in domain self-supervised pretraining for improved model robustness to label errors
- Invented 2 new Loss Functions that can identify mislabelled data, increasing label error detection rates by 13%
- Analyzed the effect of mislabelled data on accuracy and loss in common ML datasets
- Increased Deep Learning model accuracy using Confident Learning for error detection in datasets
Computer Vision Engineer, Miovision, (January - April, 2025)
- Built a Deep Learning pipeline for Semantic Segmentation of poor-visibility regions in traffic video streams
- Trained a classifier to flag frames with degraded tracking performance due to adverse visibility conditions
- Improved multi-object tracking by modeling occlusions and using temporal cues to track occluded objects
Machine Learning Engineer, Northern Digital Inc., (September - December, 2024)
- Architected a Gaussian-Splat reconstruction pipeline to generate photorealistic 3D models from 2D images
- Developed a custom Structure-from-Motion workflow for real time Time-of-Flight point-cloud registration
- Achieved sub-millimeter 6-DoF tracking accuracy by optimizing pose estimation through real-time CV filters
- Implemented a CNN-based anomaly detector to monitor and flag drift in medical-device tracking accuracy
Backend Engineer, Magna International, (April - August, 2024)
- Developed a scalable TCP data processing framework and SQL database for machine state tracking, resulting in the collection of 750 million data samples per day and annual savings of $350k
- Optimized server for high-volume data using asynchronous connections to handle concurrent clients
- Integrated AI processing using cloud computing for predictive analytics and downtime prevention
Software Engineer, Toyota Motor Manufacturing, (September - December, 2023)
- Programmed a machine vision system to control a 6-axis robot arm, automating 3 manual positions
- Developed machine learning-based vision systems for defect detection, improving accuracy by 18%
- Decreased equipment downtime by 40% through root cause analysis and reprogramming machinery
- Automated audio verification processes using Digital Signal Processing, automating 2 manual positions