09/18 2014

The 3rd Generation of Kngine

We created Kngine to allow our users to ask a question and get accurate answers quickly. Starting with this end goal, I’m pleased to announce the third generation of Kngine.


Not only does Kngine 3.0 display a ~10% improvement in accuracy compared to our second generation, which launched last year, but it is also 10% faster and more reliable. Kngine 3.0 now scores 80% on the standard TREC question answering benchmark, and 81% on our internal benchmark. That means that Kngine answers correctly 2x to 3x more questions than competitors. It is also 2 times faster. For instance, Wolfram | Alpha answers about 40% of TREC and it’s average response time is 3.5 seconds per question, while Kngine answers 80% and our average response time is 1.5 seconds.


The only Question Answering System that is more powerful than Kngine is IBM Watson, however that comes at a price; it consumes more power than 20 houses and requires a $3 million worth of a super computer to run it! And it’s only ~10% more accurate than Kngine!


But we plan to beat IBM Watson, join us and let’s make this happens together!


05/14 2013
We just launched Kngine Web beta

We just launched Kngine Web beta

04/18 2013

Kngine Summer of Code

Kngine team is happy to announce ‘Kngine Summer of Code’. 

Kngine Summer of Code is new kind of internship, where everyone who are lucky enough to join, work on one major specific project (plus minor tasks). Also the acceptance process is quite different and efficient.  This year we will have three main tracks:


Information Retrieval + Lucene:

This track is all about IR, you will implement new Information Retrieval Models and Learn to Rank Algorithms, try new news and implement them inside Lucene and evaluate your work for Question Answering.

Prerequisite: You have to be very good in Math, Linear Algebra, Java and Lucene. Good at Machine Learning. You have to read: Introduction to Information Retrieval and know Lucene architecture and source code well. Watch this playlist is big plus



This track is all about adding sugar to Java. In this track we aim to build a Java compiler (Java to byte-code compiler) that will increase the productivity by adding features like: auto/var, Pass by reference, Optional parameter, single line function, spwn & sync keywords  and others. Also we aim to improve the experience by add new classes’ libraries, such as: RE2 for faster regular expression, better collection, easy to use Data/Task Parallel library, etc.

Prerequisite: You have to be very good in Math, Java virtual machine, Java byte code and expert in Java. You have to watch Stanford Compiler course and read Basics of Compiler Design book. Also you should read about the following topics: How var works in C#, Type inferences, how Scala implemented long tail optimization, how lambda expression implemented in Java 8 and C#.


Query Parsing:

This track is all about how to parse natural language queries/questions such as: ‘Who are the employees that took more than 20 days vacation last year, non-stop flight from DC to LA that have wireless’ and convert them to logical form to generate a SQL database query.

Prerequisite: You have to be very good in Math, C#. You should take Stanford NLP course and read Speech and Language Processing, or Foundations of Statistical Natural Language Processing.


So, how it works:

The application period runs from April 18 to May 8 2013 but don’t delay. Then we will set an exam and an interview date expected to be in late May or early June (when you’re ready with the prerequisite). Yet you have to be ready by the exam/interview time, Take the courses, read the books and improve your skills in the prerequisite areas and be precise

You can apply by sending an email to ‘Jobs@Kngine.com’ with title ‘Kngine Summer of Code’  and include the track you want to work on. This program for graduates and undergraduates.

What do we mean by ‘be precise’:

If you don’t know how final or volatile works in Java and what is the difference between synchronized and lock, so you can’t write a compiler. If you don’t know how Lucene works and how BM25 model works  you can’t be in IR track.

What if I want to apply, but I don’t have everything in the prerequisite list:

You can start study from today, you will have over a month.

04/15 2013

Kngine 2.0 

We’re pleased to announce the second generation of Kngine.


While keep Kngine fast and efficient, we improved the quality of the answers and the number of questions that Kngine can answer. In our 3rd party benchmark the new version of Kngine scored 72% compared to about 25% for Siri, Wolfram | Alpha, Evi and others.


In this version we simply improved everything, from how Kngine understand the questions, answer them and format the answer. In additional to that in the last week we improved how Kngine deal with simple questions, such as: Unit and Currency Conversation, Equation, etc.


If you like to know more about our QA API that can power your Application, drop us email at Business@Kngine.com

11/30 2012
09/08 2012

Kngine beats Siri, Google Now, Evi, Skyvi, Malubba and Ziggy by 2X

Recently we were asked to compare Kngine against the Question Answering/Search Application in the market such as: Siri, Evi, Skyvi, Google Now, Malubba and Ziggy.

The 3rd party benchmark was made by an independent consultant, based on NIST guidelines and questions.

As we expected, Kngine beat all the competitors by huge factor.

Benchmark Chart

Thumbs up for Kngine and our engineers.

For more information please contact Haytham ElFadeel, CEO : HFadeel@Kngine.com

08/16 2012

06/18 2012

Updates on Kngine βeta

Here we go! We just released a new and more stable version of the app for both Android and iPhone, so download it today or update yours and tell your friends about it.

We fixed almost all the submitted bugs in the engine and the App but definitely more is coming.

Also we would like to announce our global “Local Search”:

Most of the personal assistants that understand natural language don’t have global “local search” that includes Siri, Evi, Skyvi, etc But at Kngine we always believe in challenge the status que so from first day and we’re working hard to have a global “local search” and we keep improving it. Today we just rolled out important improvements on our local search, better accuracy and links to view the restaurant’s menu for more than 500,000 restaurants around the world and even make an online reservation for about 25,000 restaurants via OpenTable.com

Honestly provide menus and make online reservation was a requested feature from you and it’s available now.

But this is not everything, we plan for other improvements in the coming weeks includes, telling if this store/restaurant/cafe is currently open or not (the working hours). Also next month we will announce significant improvements in our question answering engine and a beautiful version of the App.

06/02 2012

Say hi to Kngine βeta

We have a problem. There are about 600 Million smartphone users who suffer from using their phones to search, know stuff or get answers for their questions. We realized that search on smartphones is a pain in the a**, while it should be smart like the smartphone. So we decided to change that.

Kngine is a brand new App that answers your questions, Just ask by voice or type When is the next Cal-train, H, Revenue of GE, how to make a cheesecake, where is the nearest cvs, what the top box office for this week, etc. and he’ll get you the answer right away.

Our goal is to simplify searching for answer, information and facts; and change the usual search results from billions of links to customized answers; 

We’d love to hear your feedback, so send any comments/suggestions/complaints to Feedback@Kngine.com

Kngine is free, and the βeta is available in the Apple App Store and Android Store today!