I participated in a research team specializing in Machine Learning techniques for learning tasks in challenging data environments, such as limited or complicated data situations. My contributions spanned two primary projects:

1. Sperm Head Crop/Classification: This project focused on accurately cropping and classifying sperm head images to aid in reproductive health research.
2. Cartilage Crop out of 3D Volumes: This project aimed to develop methods for precisely cropping cartilage from 3D medical imaging volumes, facilitating advancements in medical diagnostics and treatment planning.

In these projects, I played a pivotal role in designing and implementing machine learning pipelines that handled everything from data annotation to model training and fine-tuning. The goal was to create state-of-the-art models capable of performing complex image analysis tasks with high accuracy.

---

Responsibilities

• Developed end-to-end machine learning pipelines, including data annotation, model training, and fine-tuning, to achieve high-performance models for specific tasks.
• Applied various advanced machine learning models, such as UNETR and Conditional UNETR, to address the unique challenges of each project.
• Identified and resolved bottlenecks in the data annotation process, which was a critical step in improving the overall workflow and efficiency.
• Created a custom tool leveraging Meta's SAM (Segment Anything Model) to automate the segmentation, cropping, segment editing, and labeling of images, drastically reducing the time and effort required for manual data annotation.
• Collaborated with a multidisciplinary team to align project goals, methodologies, and deliverables, ensuring successful project outcomes.
• Conducted experiments and analyzed results to iteratively improve model performance and adapt to new challenges encountered during the research.
• Documented processes, findings, and model performance metrics to contribute to the team’s knowledge base and support future research efforts.

Skills

Repo

---