A Few Observations for Developing and Marketing Successful Machine Learning Applications


(B) As I have been working on two machine learning projects for the last two years: one for an intelligent assistant application, and the other one for a predictive analytics application for the Internet of Things, I have been interesting to learn more about how other product teams were developing and marketing successful machine learning applications. And so following are a few observations…

Google (ads, search, OK Google), Netflix (movie recommendations), Facebook (face detection, Messenger), and Apple (Siri) have integrated into their back-end services, for over a few years, machine learning techniques for their consumer application offerings.

Microsoft has added early on predictive analytics to its PaaS/IaaS platform Azure. IBM is following up with Spark for its Bluemix platform, and Amazon with a machine learning service for its Amazon Web Services (AWS) platform.

SAS has offered statistical and data mining solutions to enterprises since 1976.

Many new ventures are jumping into machine learning either by:

  • Re-engineering existing products using machine learning
  • Creating new products using machine learning
  • Providing machine learning as a service (MLaaS) or a product


Creating Unique Intellectual Property (IP)

Probably few start-ups, with the exception of Ayasdi, are leveraging unique research. Ayasdi’s Topological Data Analysis (TDA), which finds insights from the topology of the data, is mostly based on Stanford Professor Gunnar Carlsson’s research for the National Science Foundation and Darpa.

It is very likely that most start-ups will create intellectual property by applying classical algorithms and approaches to new applications, or modifying classical algorithms and approaches to better match the requirements of the scope of applications.

Re-Engineering Existing Market Segments and Product Lines

Many start-ups are re-engineering existing product lines using machine learning such as:

  • Wise.io for customer service: routing tickets, recommending responses, and automated replies
  • Clari and Leadspace for sales: Clari specializing in sales forecast; Leadspace specializing in lead generation

Salesforce.com is or will likely add some of those functionalities into its Marketing Cloud offering.

Besides sales, marketing, and customer support, machine learning is starting to be integrated into information security, and in particular new threat intelligence products such as:

  • Open source MLSec project
  • Caspida (acquired by Splunk) and others for security risk analytics

Fraud prevention has always been a good segment to re-invent products.

Among start-ups that are using machine learning to re-engineer existing products, a list will not be complete without Viv, ex-SRI International and Apple employees, which is building another Siri.


Creating New Market Segments and Product Lines

Few start-ups are creating new product lines – but a few do such as:

  • Monitor 360 which analyzes business narratives using machine learning


Machine Learning as a Service (MLaaS) or a Product (MLaaP)

A few start-ups are providing the use of machine learning algorithms as a cloud service or a product:

  • Palantir for enterprises with its Gotham and Metropolis platform for vertical enterprise applications (Healthcare, Insurance, Financial…)
  • Alpine Data Labs for enterprises which need an off-the-shelf data analytics platform
  • Ayasdi for enterprises which need a platform to explore the use of TDA for their data
  • MetaMind for enterprises for image classifiers and sentiment analysis likely to expand to other application domains of neural networks
  • BigML for creating sources, datasets, models, and predictions


Note: The picture above is a Citroën car.

Copyright © 2005-2015 by Serge-Paul Carrasco. All rights reserved.
Contact Us: asvinsider at gmail dot com.

Categories: Machine Learning