Want to make money in machine learning? Find your users first

Two very different companies announced funding from two very different sources on Monday, but those companies have one big thing in common: they used machine-learning technologies to build very targeted applications.

Financially speaking, the bigger deal is the $15 million investment round that Goldman Sachs led for financial-data startup Kensho. Aside from putting cash into the company, the bank “has entered a strategic partnership to use Kensho’s real-time statistical computing and analytics technology across the firm,” according to a press release.

Kensho launched in January of this year along with $10 million from a solid group of investors. It claims its technology lets bankers and analysts ask complex, natural-language queries about financial markets and get results in real time.

The other deal is a $1 million grant from the National Institute on Drug Abuse to a company called AiCure, which will study the effectiveness of facial and gesture-recognition on improving adherence rates for prescriptions among drug addicts in a rehabilitation facility. AiCure has developed a smartphone app that uses computer-vision algorithms to determine whether the person taking a pill is (a) the person who’s supposed to be taking it, (b) taking the right dose and (c) actually swallowing it.

[protected-iframe id=”73c6fb3ebfe4c3c35c6d42d23bd6a602-14960843-6578147″ info=”//player.vimeo.com/video/104977764″ width=”500″ height=”281″ frameborder=”0″ webkitallowfullscreen=”” mozallowfullscreen=”” allowfullscreen=””]

AiCure was founded in 2009 and has so far garnered $3.4 million in grants from the National Institutes of Health, as well as revenue from “major” pharmaceutical companies, founder and CEO Adam Hanina told me. Those companies use the AiCure app to monitor participants in phase 2 clinical trials, which often take place in outpatient settings and have historically relied on participants’ word in order to assess whether they followed the guidelines. Miscalculations on actual adherence during the trial phase can cost lives, and lots of money, once a drug reaches the public.

Hanina said AiCure could potentially work with any institution, including insurance companies and court systems, that might have financial or other reasons for wanting to ensure people are following the directions on their prescriptions.

That Kensho and AiCure are apparently having success (or are at least able to raise money) in some specific fields is part of a broader trend in the machine learning space. Whether it’s deep learning models or more-traditional clustering algorithms, investors and paying users generally appear much more excited about applications than they do about general-purpose software.

The gap might well narrow as familiarity with certain techniques grows, and they become easier and better integrated with companies’ existing big data platforms. But for now folks might prefer machine learning experts give them an image-recognition service or predictive systems-management tool rather than teach them to build such things themselves.