How do I validate that a robot is safe and reliable?
You can run a test, which is the starting skill, so you are further along than you feel. Validating true safety and reliability defeats you because it is a bigger question than any single test answers. A test shows the robot worked once, under known conditions. Safety is the harder claim that it will keep working, and fail safely, across the conditions you cannot fully predict. The shift is from proving it works to trying to prove it does not. Ask what would make this dangerous, and go looking for that on purpose. Reliability is built by asking not just does it do the right thing, but what does it do when things go wrong: when a sensor lies, when a part fails, when the input is nothing you planned for. Break it in the lab so it does not break in the world, because a failure you find yourself is a gift and one someone else finds is a disaster. Start with the single worst thing that could happen if this robot failed around a person, and build your validation outward from there. Safety work is never truly finished, and making peace with that is part of doing it well. You are not failing at validation. You are meeting how genuinely hard and how genuinely important it is, and taking it seriously is the first real qualification for the job.