Fine AI

Home /Fine AI

Crawler

Our distributed and multi-tenancy crawler has following features

  • Auto Proxy Rotation
  • Automatic Alerts
  • Configurable Parer API
  • Ability to restore crawl status from failure {E.g. Network Outages etc}
  • Web based control interface for:
    1. Configuring Crawl Jobs
    2. Reporting
    3. Ability to set level of alerts {E.g. %of Failure etc}

Matching Engine

Tinder, Google AdWords and Pandora are some examples of businesses whose core is based on matching engine. Matching elements from one set to another has lot of business value, lets consider a few of them:

  • Matching resources to jobs
  • Matching a product to a potential customer
  • Matching customer service representative to a complain customer might have We've built few APIs that allow building quick classifiers for Textual and Image datasets.

Few scenarios where this is used is:

  • Document Classification
  • Anomaly Detection
  • Matching Entities

Custom Search Plugins

Search is an important differentiator for most online businesses. In computer science it is an active area of research. To make improvements on search engine key focus areas include:

  • Understanding users intent
  • Making search engine resilient to typos
  • Creating a ranking model that aligns Business and Users preferences
  • Giving structured response to ambiguous queries

Lets say a user types in

  • blue checked shirt
  • blue checked shirts for men
  • blue checked shirts for women less than 2000

The intent in these three cases is different, a traditional keyword based search would have treated all the three results almost the same. Our custom implementation of search engine handled this issue by identifying following attributed in a query

  • Product Category
  • Color
  • Special Attributes {E.g. 4g compatible mobiles, 24gb storage phones}
  • Gender
  • Price Quantification {including ranges E.g. between 300 and 500}
  • Brand

Users will not always type in the right query, including the correct spelling the onus of figuring out what the user meant is huge, let look at some of cases that a typical search engines have to handle:

  • Typos: {E.g. bleu shirt when they actually meant blue shirt}
  • Synonyms {E.g. black blazer and black suit}
  • Specific model {E.g. Moto X Play}
  • Query Segment at wrong place {E.g. samsung suit case as opposed to samsung suitcase}
  • Dealing with abbreviated units {E.g. GB, Gigabytes}

These issues are generally taken care of with special run-time string transformations that make a query more consistent to indexing scheme.

Ranking is a core component that needs to get right, it needs to be balanced in a sense Business Objective and User Experience Objectives are met. Lets say as part of key partnership Business might want to promote a brand even when the user is looking to purchase competing brand. How do you rank products in that case?

Recommendation Engine

Recommendation engine are built to improve user experience, allow people to discover items that they may be interested in and increase overall retention. This proposal written talks about building a generic recommendation engine that can be deployed in multiple scenarios. Some of the settings in which we envision the recommendation engine to work include

  • New Recommendations
  • Jobs Recommendations
  • Music Recommendation
  • Customized / Segmented Email Marketing

General algorithm for building and testing recommendation engine include

  • Collect Data (this is optional, in case of 1. content based we skip this step)
  • Select Similarity Metric
  • Select Recommendation Model
  • Evaluate and test Recommendation Model

Each domain in which recommendation engine is deployed variety of data and features can be exploited. This proposal talks about different classes of recommendation engines and approaches that we will be taking in implementing them.There are 3 main classes of recommendation engines that are used, which include:

  • Content Based: These family of recommendation algorithms don’t need rating information to work with. They use purely content features, this is good for the fact that it helps the Cold Start problem associated with recommendation engines.
  • Collaborative Filtering: These family of algorithms do use rating information. Basic idea is to predict user’s preference given their own historical preference as well other user’s rating information.Depending on how you implement them it can further be classified as Item Based or User Based
  • Hybrid: These are the classes of recommendation algorithms that use both Content and Collaborative filtering based approach. Example of a hybrid is one built by NY Times

Churn Analysis Engine

Churn is defined as number of people who have ceased to do business with a Company over a fixed period of time. Businesses which utilizes network effects are hit heavily both in terms of lost revenue and spent marketing efforts.

Sectors like Telecom, Retail, Hospitality and Subscription based services are few sectors where Customer churn is taken seriously. Being able to track Churn and predict it implications of:

  • Improved user experience and thus revenues
  • Optimization of Marketing expense

Churn Analysis is collection of activities that including collection of data from CRMs to process improvement across Operations and Marketing, which will reduce number of customers leaving for competition. If we're able to identify and predict dis-satisfied customer we can take corrective actions.