Hey Product Hunt! I'm Alex, co-founder of Graybox. We're a team of AI engineers building what we think is a missing piece of the model-training stack. When you train a model, most of what's actually happening, how each sample behaves, where the model struggles, and which data is hurting you, stays invisible until a run is over. WeightsLab captures granular, per-sample signals during training, so you can see what's going on and make adjustments on the fly rather than finding out hours or days later. Today we focus on perception systems and the data behind them (images, video, LiDAR). But the longer-term vision is full end-to-end traceability, connecting the patterns a model learns back to the exact data that produced them, across all model and data types. We'd genuinely love your feedback, especially if you've ever wasted experiments because of a data problem you caught too late.

Weightslab

A tool to develop models between BlackBox and Whitebox

A tool to develop models between BlackBox and Whitebox