Our data science offering allows the enterprise to dig in to its data, be it of any type and of any size, to extract the insights required to power its businesses. The service is designed to help equally the small business looking to make the most from its exclusive users as much as the big enterprise that has millions of customers spread across markets. This service is powered by our knowledge and experience in designing and implementing massive data processing and analytic systems.

The first step in our data science offering is to chart out what data exists in the organization today and how that impacts the business outcomes that are critical for the enterprise. One of the deliverables from this step is the identification of gaps in either the data itself or the knowledge of the data as required for key business outcomes.

In the second step, we carry out an assessment on whether the data required for filling in the gaps is of good quality and has enough richness. With the outcomes of this assessment, both the enterprise and we are well placed to understand what analysis can be carried out and what additional data sets or flavors are required to enhance any derivable value.

The third step is the act of physically executing the data science portion of the offering. Here, our trained data scientists using various tools to dig into the data either onsite or offsite and provide immediate feedback.

We pride ourselves in following an agile model of delivery, where there is constant interaction with the in-house experts from the enterprise as and when new insights are found so that new hypotheses are arrived at and tested in a rapid iterative manner.The choice of tools including for data munging, analysis and visualization is typically provided by the enterprise to conform to corporate standards.Alternatively, we are happy to bring our own toolkits to the workplace and can provide our recommendations too.

Our Industry View

Recommender Systems & Approaches

Recommender systems are more ubiquitous in today's world than is apparent, and in industries ranging from e-commerce to telecom and banking, there are a wide range of use cases. In this post, we explore the different approaches to recommender systems, and how their implementation varies based on context. In addition to a discussion on technical aspects and algorithms, we also discuss the impact on behavior and the balancing and reinforcing loops that recommender systems can impact in consumers.

The 1-hop neighbours problem

The problem of computing number of friends connecting two individuals forms the heart of people recommendation engines. This problem is easy to express in small data contexts (e.g. using SQL) but are notoriously hard to solve with good performance on massive, real-world networks. We discuss why.