Blazing Customs / Research
Research & new products
Work done to find out whether something works before anyone pays for it at scale: hardware and embedded prototypes, research into how models behave, and new physical products. Most of it ends in a written answer, and sometimes the answer is no.
Embedded and robotics work taken from a simulator to a physical prototype: flight control and autonomy on drone airframes, mapping and sensing, and an optical channel that carries data from a screen to a camera with no network between them. Measured on the bench, with the failures written down.
Typical work: A feasibility build before you fund the real one; sensing and control on a drone or robot platform.
Research into how models behave on the inside: adjusting a behaviour directly in the weights instead of pleading with a prompt, how much of a model survives compression, and how to tell a genuine improvement from an artefact of the test. A lot of it ends in a written-down negative result. That's how a client gets told no before paying for the wrong thing.
Typical work: A go / no-go on a model idea; a test harness that can't flatter the result.
Product development that starts before there is a product: sourcing, prototyping, and the fabrication workflow for printed parts, alongside modelling work that predicts how a material will smell from its chemical structure. Early-stage by nature, and described here as that.
Typical work: A prototype run; a sourcing and feasibility check; a model of a property that's slow to test by hand.