Brad Porter

I wouldn't just read books. Books are good for foundational background, but are already too far behind. Find some of the top influencers in robotics on LinkedIn and follow them. Also, follow the research coming out of places like MIT, Stanford, CMU and Berkeley.

It's important to think about how the public will perceive your robot. We want robots to be socially accepted. Scary images of robot armies or robots wielding weapons or flame throwers accomplish just the opposite. When you're about to release that video, ask yourself "would an 8 year old be excited or scared of this? how about an 80 year old? how about your next door neighbor?"

This might seem odd, but I find ChatGPT is a great partner if you're feeling any level of imposter syndrome. ChatGPT can help you get up to speed in a space quickly with the basics, but also, it's equally reassuring when you realize you have more expertise than the collected knowledge of the web encapsulated in model weights. You can also go to Youtube and search for the 2015 Robotics Challenge and watch the experts in the field struggle with their robots. Robotics is hard. Just keep building.

Leading a robotics team is particularly challenging as you need to be able to hire and develop individuals with very different skillsets, from mechanical to electrical to software and autonomy. It helps to be an experienced roboticist who has shipped and delivered robots and can look across the entire system, while identifying those domain experts who can bring their expertise to the systems challenge.

There are three sources of good scoping data. One is the intuition of an expert who has done a number of these projects before. The second is a rigorous bottoms up work/effort analysis. The third is a model based on past performance of similar projects. In the ideal world, you combine all three and reconcile any differences. To do this, you need to start building a heuristics driven model as soon as you have 3-4 projects under your belt. You need experienced experts and you need a program driver to get the bottoms up estimate from the team. But if you have these you can do a very accurate estimate in a couple of days. If you don’t have these, then asking an expert who has done a few is the fastest.

Process optimization isn’t independent from business objectives. An inefficient process that isn’t impeding the business and for which the cost of doing it better doesn’t pay back anytime soon, is a fine process to leave alone. Engineers and operators hate waste and inefficiency, but sometimes inefficient but working well is what the business needs. If you’re successful to the point that this process becomes so inefficient that it’s expensive or a bottleneck, fix it then. But until then, it’s ok to not be perfect at everything. I recall one of Jeff Bezos’s lines was “imagine how much more we’ll love this business when we get good at it.” It’s ok not to be great right away, as long as you’re good enough that the business works.