From the noise to the useful signal
Home Careers Blog ← Back to Blog From the noise to the useful signal Paweł Budzianowski, Kobi Kelemen · February 25, 2026 Scaling up data collection is a feat in itself, regardless of how you collect it. At Lute we collect the data around the clock in different environments and with different embodiments. Every piece of hardware has its own quirks, every teleoperator has a unique style of movement, and every data annotator has their own taste in labelling. When you add general videos of human actions collected during pre-training, you amass a large amount of data every day — but with high noise. In this blog post, we explore how we are building a set of quality metrics to keep improving data quality and the performance of final policies on challenging real-world problems tackled alongside our manufacturing partners. Tackling real-world problems - collaboration with iBombo iBombo is a global leader in bike repair and solar charging station production with over 24000 stations deployed ac
Home Careers Blog ← Back to Blog From the noise to the useful signal Paweł Budzianowski, Kobi Kelemen · February 25, 2026 Scaling up data collection is a feat in itself, regardless of how you collect it. At Lute we collect the data around the clock in different environments and with different embodiments. Every piece of hardware has its own quirks, every teleoperator has a unique style of movement, and every data annotator has their own taste in labelling. When you add general videos of human actions collected during pre-training, you amass a large amount of data every day — but with high nois
Explore this link on the map →