Saeid Taleghani

Experience

• 

  Rayan Naqsheh Consulting Engineers

  Nov 2016 - Sept 2018

  Remote Sensing Expert

  • Preprocessing aerial and satellite images.• Producing orthogonal images for drawing two-dimensional maps.• Producing 3D models using sequence images for drawing objects in 3D form.• Working in Research and Development department.

• 

  University of Waterloo

  Apr 2019 - Jul 2023

  - Designed a conditional random field (CRF) model to detect oil spill candidates by utilizing spatial information and statistical properties of compact polarimetric SAR data.- Designed a feature similarity by considering the polarimetric properties of compact polarimetric SAR data.- Evaluated three well-known optimization methods so that the resulting CRF model can be tailored to the characteristics of compact polarimetric SAR data.- Outperformed conventional Markov random field and CRF models. Show less

  • PhD in Systems Design Engineering, University of Waterloo

    Sept 2018 - Jul 2023

  • Research Project: Designing a New Vision Transformer to Classify Land Covers

    Mar 2022 - Mar 2023

  • Research Project: Region-Based CNN Sea Ice Classification

    Mar 2021 - Mar 2022

  • Research Project: Oil Spill Candidate Detection

    Apr 2019 - Mar 2021
• 

  NextGen Environmental Research Inc.

  Sept 2023 - Aug 2024

  Machine Learning Engineer

  1- Coded end-to-end machine learning projects, overseeing data preprocessing, model training, and controlling performance of the trained model for real-world lake ice hazard and flood detection applications.2- Developed and executed data collection plans including ground truth generation, normalization, and transformation.3- Trained and evaluated ML models on unstructured data and assessed the performance of various models through quantitative and qualitative comparisons to achieve optimal results.4- Leveraged distributed GPU computing frameworks (PyTorch lightning) to accelerate convergence and enhance efficiency, reducing training times by up to 50% while enabling large-scale, complex model training.5- Deployed models and workflows on a high-performance local server, leveraging GitHub for version control and server-side execution.6- Created comprehensive reports and visualizations using tools such as Matplotlib and Seaborn to illustrate key performance metrics and communicated data insights, model performance, and project outcomes through technical presentations. Show less