Kushal Shah

Master’s graduate, specializing in Data Engineering, Computer Vision, Machine Learning, and Software Development, with almost 3 years of relevant experience, seeking career opportunities at the intersection of these fields.

Key Achievement: Implemented a self-supervised MoCoV2 model with a CI/CD pipeline, boosting F1-score by 56.52%; while increasing data handling efficiency of over 300,000 unlabeled images.

Geberit Gmbh 03/2024 – Current

Machine learning Researcher Pfullendorf, Baden-W¨urttemberg

• Spearheading the development of merging and refactoring strategies for machine learning pipelines, notably

MoCoV2, enhancing modularization, encapsulation, and operational efficiency.

• Code review of peers and giving detailed feedback.

Geberit Gmbh 06/2023 – 02/2024

Machine learning Researcher - Thesis (Grade: 1.3) Pfullendorf, Baden-W¨urttemberg

Thesis Title - ”Self-supervised pre-training for hierarchical image classification: A study on unlabeled data”

• Research Focus: Investigated the efficacy of self-supervised pre-training in hierarchical image classification, with a focus on machine learning and computer vision techniques.
• Data Optimization: Streamlined data management by pre-processing over 300,000 unlabeled images. Simplified

ETL by creating a user-defined function in Azure Cosmos DB for filtering, boosting data handling efficiency.

• Pre-training Model: Developed and integrated the MoCoV2 model in Azure by building a CI/CD pipeline for automated deployment and self-supervised image embeddings extraction.
• Fine-Tuning: Prepared a classification head using transfer learning principles with MoCoV2 as the backbone,

improving precision (33.33%), recall (46.00%), and F1-score (56.52%) over traditional supervised methods.