I’m excited to announce our new Bayesian consultancy focused on PyMC3: PyMC Labs. We work with companies ranging from startups to Fortune 500.
Most of our clients have a PyMC3 model that might be slow, not sample efficiently, or can be extended in interesting ways (e.g. making it hierarchical). We also develop new models from scratch to solve applied business problems.
In addition, we also do corporate workshops where we teach PyMC3 in a hands-on way.
I have also worked in consultancy for a while. Many times have I tried to promote a probabilistic programming approach to business problems. Most often, the fear of them having to maintain it after we left was one of the main arguments to stick to “standard” ML solutions. The problems were also often of a nature that required only prediction, without uncertainty estimates (as long as the accuracy measures were still on point!). Once, I have indeed used PyMC3 for a project (very heavy towards inference, in the oil and gas industry, where we estimated effects of drilling and maintenance on the production of existing wells). It was great fun and very successful! In that consulting company, not many consultants were in fact experienced with PPL and many seemed surprisingly reluctant to get familiar with it.
In my current work, in the water utility industry, I’m slowly pushing towards PPL usage as well and we will soon start our first trial project (predicting the inflow of objections to water tax assessments). I think my experience and willingness to figure it out will be enough to push it through, otherwise I know where to find you! Also, I’ll suggest PyMC Labs and your stellar team to whomever may need help with their probabilistic modeling efforts. Do you serve customers worldwide (and therefore perhaps remotely)?