Case Study: Tiny Mile achieves 19.5% higher model accuracy and 32% lower retraining costs with Activeloop

How Tiny Mile, Manot, & Activeloop Increased Accuracy, Reduced ML Retraining Costs, & Streamlined Robot Delivery with Data-Centric AI

Tiny Mile, a company specializing in last-mile delivery using autonomous robots named Geoffrey, faced challenges in ensuring their robots' reliability. Their computer vision models struggled with unexpected real-world scenarios and inaccurate object detection, while an inefficient data pipeline and lengthy feedback loop slowed development and increased costs. To address these issues, they partnered with the vendor Activeloop.

Activeloop, alongside Manot, implemented a solution using its Deep Lake data platform to establish a continuous feedback loop and an efficient data pipeline for model retraining. This allowed Tiny Mile to quickly curate data, retrain models, and improve accuracy. The results included a 32% reduction in retraining costs, a 19.5% increase in model accuracy, and a 10x improvement in time to production.

Tiny Mile

Ignacio Tartavull

Chief Executive Officer