Late Night Ramblings

Just some late night ramblings.  I thought it would be fun to create a category on the new blog called “late nite ramblings”. These are the ideas that form somewhere between coherence and “maybe I shouldn’t post it”.  But it’s often those short-lived spikes of semi-consciousness that lead to the most discussion and provide the best fodder for disruptive theories. It’s either that or I’m feeling a bit tired. Feel free to call me an idiot.

At IM, we love what what we do.  We love to help people develop a clear understanding of the possibilities and challenges of modern prediction and machine learning techniques from a business perspective. We won’t sell you something you don’t need, and we’ll be honest about it. We are proud of our lean, no-nonsense organization. Those of us going out on “sales calls” are the same people who have their sleeves rolled up during the analysis & development stages of a project. We like to think that we are all technically savvy enough to plug-in to many-a-scrum-team and be effective right out of the gate.

And although we can be lean in many ways, there are some corners we simply can’t cut. Not if we want to deliver a really effective solution. Not if we want the kind of solution that an in-house team would be expected to deliver. We could take shortcuts but then we would be a “black box” solution company – those already exist. But the crux is in iteration. Companies that understand this nuance are trained to think that this level of attention is only feasible economically if undertaken by an internal team. After all, you don’t have to pay them overtime.

And herein lies the challenge for us. The best engagement  is with a company that wants  more than “check-box” “feature” functionality. The best potential clients are also those who are most likely to build an internal team to take on the challenge. They want to develop a solution that resonates with their audience, that embodies their value proposition – a solution that is “core” to the very reason the site exists. Again, it’s the kind of solution most businesses only expect to get out of internal teams, not vendors, unless you have a bottomless wallet.  Then call in IBM!   Yet we still strive to offer a consulting solution that comes as close to that “in-house” experience as possible within a reasonable cost. We think this is the future. Partly because we like to work like this. And partly because we realize that demand for our skills will continue to greatly outweigh the supply. To be upfront about it – it might take you one year to put together an effective in-house team to deal with the business problems at hand, orrr it might take us 3 months to develop a solution to your problem. We provide the launch pad.  There is always some trade-off.

And I might mention that we have seen our share of “big data” teams drinking their own Kool-Aid, especially in “start-up” environments.  Those guys have been thrown to the ML Wolves with real-world data or business experience.  But we’re confident they can eventually figure it out, they’re super-smart people.   But when the top talent CS / ML graduates get swept away by well-funded bubble companies, what will you do?

I issue a challenge to businesses to rethink the traditional vendor relationship when it comes to machine learning & prediction solutions. If you want to take advantage of some of the better talent that is out on the market right now, the type of talent that doesn’t want to become “in-house talent”, you will have to challenge the out-dated processes that have become accepted practices at your company. You, like us, will be forced to think outside-the-box to come up with solutions to age-old problems, and find ways to convince your executives that a fancy powerpoint wearing a suit is not the solution to your problems. [No surprise, by the way, that most companies who face a brick wall here are also those companies who are struggling for relevance in today's WWW-augmented business reality]

One more thing …  if you’re lucky to work with top machine learning solution providers, be sure to have your team interact with them as much as possible. The more you can do to expose them  to those with practical experience in ML, the better off you will be. Trust me, your engineering team will thank you for it. Tomorrow, CS engineers will be pouring out of college with basic ML chops. And newly minted ML majors will be graduating with basic programming chops. Neither will be qualified as a “machine learning engineer”, although both will say they are one…

Or … you might think I’m crazy…