Here is an interview with Andy Bartley, CEO of Algorithms.io . Algorithms.io has been selected as one of 30 start ups to participate in this year's SV Forum Launch: Silicon Valley conference. In this wide ranging interview Andy talks of algorithms, APIs and the emerging and lucrative field for Big Data analytical algorithms.
PW- What is Algorithms.io all about? How did you come across creating, developing and nurturing your start up idea?
Andy- We believe applications need to be more intelligent.
In a world with millions of applications and ubiquitous data, what will set apps apart is how they use data to make proactive or predictive decisions. Algorithms are one of the components of accomplishing this goal. We’ve been using algorithms for predictive applications such as product recommendations for years. Every time we wanted to find a new algorithm it required hours of online search in various repos and libraries, and digging through academic white papers. We felt there had to be a better way to integrate algorithms into new applications.
Algorithms.io is starting to solve this by providing intelligent APIs as a service. Algorithm designers and data scientists can publish and monetize their algorithms, developers can integrate intelligence into their apps, and businesses can request custom APIs and applications to enhance their data.
Andy- Apps are gathering data, but to add intelligence, it usually requires integrating a custom library or rewriting an algorithm. When a new programming language comes out with a similar algorithm, that's faster, better or cheaper, lots of code needs to be rewritten, what if you could just swap out and select a new algorithm from a catalog and apply it to your problem domain? That what were trying to provide for developers.
We created an industry-standard RESTful web service API that makes possible for easy developer adoption. Mashape provides Algorithms.io with a simple and easy to use platform that exposes our API services through their marketplace so its a natural fit.
PW- Who would be a defined segment or target customer for your API? Could you please give us a case study based on live usage? Any usage stats on API till now?
Andy- Right now, we think about our algorithms as addressing three types of business needs: predictive analytics, dynamic optimization, and social influence.
An example of each would be:
Predictive Analytics – predict customer churn
Dynamic Optimization – Content optimization on websites to increase customer conversion
Social Influence – Analyze and score social graphs to identify top influencers and high-value user types.
What we heard from our customers during our alpha was that they need more information on how to put our algorithms to work to solve business problems. Many are still struggling with data infrastructure and ETL problems. We’ll be building out more end-to-end applications based on our current customers with the goal of making it easier for developers to think of applying the algorithms to their own work.
PW- Describe your involvement with open source and R for algorithms. What do you think about Python and R in terms of algorithmic support. Any plans for Julia or Clojure or any other languages that you have considered?
Andy- Our primary focus will be to allow R and Python users to collaborate on our platform effortlessly. In our assessment, this approach will keep data scientists and big data developers in their most productive development environment and promote broad base adoption.
We started with support for R due to its popularity and our own familiarity with it. We really like R because it simplifies many difficult operations and there is a well-established open source community that we can contribute to as we roll out our platform. Python is next on the list for us.
Julia and Clojure are very interesting in their native parallel computing capabilities and fits well with our cloud backend. At this point we plan to let customer demand dictate whether or not we integrate these into our platform.
PW- What are your plans for 2013?
Andy- Our goals for 2013 are to roll out multiple language support for packaging and runtime, an improved catalog with more use cases, and custom solutions and apps that are built on top of our service. Later this year we will be hyper focused on providing a new way to glue algorithms together.
PW- Apart from the following four algorithms, what are some of the others that you are thinking of rolling out soon?
Andy- We plan to let our customers determine which algorithms we launch next. We are currently working with a limited set of customers on specific use cases around predictive analytics and social influence scoring. Based on these use cases we are identifying new algorithms to add. This list currently includes additional types of Regression Analysis, Network Link Analysis, NLP algorithms including Sentiment Analysis, and more anomaly detection.
Longer term, we are really looking to get the communities involved so they can add and hopefully monetize their algorithms by solving specific use cases. As new algorithms replace older ones or programming languages change we want to provide a service to support the transition by abstracting the interface and runtime so you are just dealing with an API.
Algorithms on demand, anyone? Just another (algorithms.io) API call away!