
Engineering blog
Kestrel Agritech gives cotton growers a forecast
Narrowing hyperparameter sweeps and keeping a sprawling field-ML project legible.
Hear from the teams who run Andesite in production — how they track, evaluate, and ship models at scale.

Narrowing hyperparameter sweeps and keeping a sprawling field-ML project legible.

Watching training in real time and catching regressions before they ship.

An ML stack assembled for efficiency, flexibility, and field performance.

A 9-month project teaching a robot hand to solve physical puzzles unscripted.

Evaluating and debugging autonomy models against 2.4M labelled clips.

One report surface keeping a 41-person distributed research team aligned.