Sukriti Paul

Blog

My thoughts and insights on computer vision, AI, and more.

The Future of Vision Tokenizers Published: March 5, 2025 Exploring how adaptive token lengths and mixed-precision schemes are revolutionizing the way we process visual information in deep learning models. Read more →	Quantization Techniques for Vision Models Published: February 20, 2025 A deep dive into post-training quantization strategies and their impact on model performance, with practical examples from my work on the Cosmos tokenizer. Read more →
Building Better Training Datasets for Vision-Language Models Published: January 15, 2025 Lessons learned from creating PixelProse, a 16M+ image dataset with synthetic captions, and its impact on multimodal model performance. Read more →	ML Approaches in Proteogenomics: Insights from nORF Research Published: December 10, 2024 How machine learning is transforming our understanding of novel proteins and what this means for the future of structural genomics. Read more →