Skip to content
Koidra

Careers

Build automation that runs a real plant

Most machine learning work ends at a metric on a slide. Ours ends with a setpoint written to a controller in a plant that is running right now, which is a much harder and much more interesting place to stop.

What the work is actually like

The problems are physical

A dryer has thermal mass. A press wears. Debugging sometimes means a phone call to someone standing next to the machine, and the feedback loop runs through a real plant rather than a metrics dashboard.

The bar is a working plant

Our control software has run a commercial crop for a full season against expert human operators, twice. That is the standard the work is held to, and it is a different standard from a demo.

Small teams, wide scope

Engineers here touch the control layer, the platform and the product. That breadth is the job rather than a temporary state.

Principles

These are the ones people actually quote at each other, which is the only test of whether a value is real.

Work backwards from customers
Start at the plant floor and reason back to the software.
Transparency
Say what is true, including when it is inconvenient.
Ownership
The person closest to a problem carries it, including into someone else’s area.
Bias for action
A decision made and revised beats a decision deferred.
Iterate and launch
Ship the version that teaches you something.
Disagree and commit
Argue hard, then move together.
Do not repeat yourself
Applies to code, to documents and to meetings.
Growth mindset
The people who get good at this are the ones who keep getting corrected.

Open roles

We hire across engineering, operations and commercial roles, in the United States and Vietnam. If you see the problem above and want to work on it, write to us with what you have built. A short note about a real thing beats a long CV.

careers@koidra.ai

More about how we work is on the about page.