Sara Naseri Golestani

Proficient in designing, training, and deploying machine learning models, including neural networks and deep learning architectures, for diverse applications such as object detection, natural language processing (NLP), and resource usage forecasting. Adept at creating efficient data processing pipelines for complex datasets, including 3D point cloud data, and extracting meaningful features for machine learning tasks. Hands-on experience in developing end-to-end ML pipelines, from data preprocessing to model deployment, utilizing Azure and Docker for scalability and efficiency.  Skilled in a range of programming languages, including Python, MATLAB, and SQL, with hands on experience in C. Proficiency extends to Flask, Java, and SQL for web service development and Restful-API implementation in ML projects.

Worked closely with cross-functional teams to identify and prioritize data requirements for machine learning projects, ensuring the availability of high-quality data.  Conducted regular code reviews and provided feedback to team members to maintain coding standards and best practices.  Collaborated with stakeholders to define project scope and deliverables, ensuring alignment with business objectives.  Presented project updates and findings to senior management and stakeholders, effectively communicating complex technical concepts in a clear and concise manner.

•  Played primary development role in data science team in developing and implementing machine learning pipeline object detection on 3D point cloud data, resulting in a 20% increase in accuracy.
• Successfully implemented GitHub actions for an ML Bootcamp project, reducing the reviewing time by more than 50%.
• Implemented automated model training and deployment workflows using tools like Docker and Azure, reducing the time required for model deployment by 50%.
• Developed novel algorithm for sentiment analysis, achieving sentiment prediction accuracy of 90%.
• Collaborated on analysis and presentation of findings for clients in 3D point cloud data project, contributing to the scientific community's understanding of the subject.

Researched and implemented machine learning algorithms to accurately forecast resource usage for the SAKKU Platform.  Developed Restful-API modules to efficiently collect required data for machine learning development. Designed and implemented file search feature to improve user navigation and productivity. Collaborated with cross-functional teams to ensure seamless integration of ML algorithms and Restful-API modules. Conducted regular testing and debugging to maintain the functionality and efficiency of the ML algorithms and Restful-API modules. Provided technical support and guidance to team members, facilitating smooth workflow and timely completion of projects.

• Successfully implemented machine learning algorithms that significantly improved the accuracy of resource usage forecasting for the SAKKU Platform.
• Streamlined data collection process by developing efficient Restful-API modules, resulting in enhanced productivity and reduced time constraints.
• Enhanced user experience by implementing a file search feature, garnering positive feedback from clients and users.

University of Calgary, Calgary: 

• Object detection with the help of Multitask Learning (MTL) to improve the Mask R-CNNs ability to detect small objects with higher accuracy. 
• This method was applied on backbone of these object detection methods to enrich their output feature maps.
• The Object detection methods used for improving the Map in object detection on UAV data were YOLO, Mask R-CNN, and DETR.

Ferdowsi University of Mashhad, MASHHAD

• Noise cancellation in magnetic-based localization using complex-valued neural networks.
• Robust magnet-based localization system to sensor dis-placement using the Genetic Algorithm (GA).
•