We publish what we learn.

Alright, let's talk about NVIDIA's GR00T N1. In our series on Demystifying robotic Foundation Models, it introduces unique approaches to data augmentation, synthetic trajectories, and leveraging action-less videos.

Demystifying robotic foundational models: from Pi0 to Gemini-Robotics & Groot N1. Deep dive into Vision-Language-Action models, flow matching, and the training recipe for generalist robots.

Building an AI system for cattle weight estimation using smartphones. Our research enables farmers and ranchers to accurately estimate the weight of cattle by taking a photo, eliminating physical weighing scales.

How we built the largest publicly available longitudinal dataset for dementia prediction, using spontaneous speech analysis to detect cognitive impairment 10-15 years before symptoms appear.

Lessons from building AVVA, an AI veterinary platform powered by multiple LLMs. Exploring fine-tuning, memory management, model selection, and the challenges of providing helpful medical advice.