about us



Identify, in real-time, potential customers signaling intent to buy a product and in that moment engage them. Zenti reaches customers through Twitter before they use search.

App Infrastructure

Generic framework and server backend allowing multiple instances to be set up. Flexible Class and Interaction set up and administration.

Social Business Insights

Real-time and historical analysis of customer (and competitor’s) sentiment regarding brand, product offering, campaign, launch events, price and service.

App Infrastructure

Generic framework and backend having the ability to spin up multiple instances. Flexible Class set up and administration. Offered as a Subscription Service or as a behind the firewall deployment.


With an average of 20 suicide related deaths per day in the US, Veteran suicide rates are at an unacceptable level. The NVF, Dr. Joe Franklin and Zenti have combined resources to tackle this epidemic of Veteran suicide, by using the best on-the-ground expertise with cutting edge psychological research and machine learning intelligence.


Dr. Joseph Franklin's work is amazing. Zenti is proud to help enable his innovative approach for the identification and prediction of suicide.

vandytaplab.com – General information about Dr. Franklin’s work

vandytaplab.com/treatments – Specific information about treatments

Zenti analyzes, classifies and determines what content is ‘about’ and then provides users with intelligent, actionable outputs.

0 Tweets Processed Per Day
0 Average Click Through Rate
0 Average Conversion Rate

Case Studies

Identifying Behavioral Change in Language: the Enron Emails 

by Steven Cracknell 

With the volume of data generated by businesses today, machine learning software is necessary to be able to handle the sheer volume of data. In this document, we demonstrate that a single classifier on obscenity would have detected a behavioral change in the language used by the Enron corporation in emails over the year leading up to the company’s dissolution.

Identifying Suicidal Behavior Among US Military Veterans and Soldiers

 by Steven Cracknell  

This case study illustrates a proof-of-concept constellation of two classifiers (military service and suicidal language) that are jointly applied to the same text samples. The ability of the Zenti system to distinguish between the language patterns indicates clear separation in the language scoring between these two constructs

Towards Classification of Factors to Improve Social Media Suicide Identification and Prevention

by Steven Cracknell and Ben Cannon 

This paper proposes a structure for a Risk Matrix of language associated with suicidal behaviors. The purpose of our Risk Matrix is to (1) understand the broader picture of suicide risk, (2) acknowledge that there are many risk factors that influence suicide, and (3) to prioritize that combind risk factors can lead to the greatest risk (Kessler, Borges, & Walters, 1999). The outline can provide the basis for training new classifiers using Zenti’s system.

Identifying Suicidal Behavior Among the General Population 

by Steven Cracknell 

This document illustrates a proof-of-concept classifier prototype that can identify suicidal language amongst Twitter users, and an ability to retrieve and score an at-risk person’s Twitter history to establish whether there is a pattern of at-risk language indicating suicidal tendencies. We highlight one specific example of a Twitter profile that had a strong pattern of language indicating suicidal tendencies, and look to understand how such identification can be transitioned to intervention.

Cyberbullying in the US

by Steven Cracknell and Antonia Czinger 

This case study illustrates the use of a constellation of classifiers, a snapshot of a proof-of-concept classifier designed by the Zenti team, and a summary of background research, which will serve as a starting point for further classifier development.

Supplementing Police Intelligence (Research Design Plan)

by Karoline Pershell 

This document illustrates one use case for a constellation of classifiers, identifying specific examples of classifier constructs and detailing how an end user would interact and take action on such data.

Development of Arabic Classifiers for Countering Violent Extremism (Research Design Plan)

by Steven Cracknell and Karoline Pershell 

This document discusses the complexities of Arabic dialects, and discusses a strategy for building a constellation of meaningful classifiers that would aid in countering violent extremism.

our team

John Thornton

Co-Chair, Brookings Institution; Executive Chairman, Barrick Gold; Chairman, PineBridge Investments; Professor, Tsinghua University; former Co-President, Goldman Sachs

Steven Cracknell

Extensive experience leading teams, commercializing innovations and managing global projects. Serial Innovator of solutions enabling accessible systems for diverse communities. Product Development expert. Goldman Sachs, Thomson Reuters

Ben Drees

Co-creator of Freebase which became the kernel of Google Knowledge Graph. Implementor of the award winning accessibility program “Outspoken for Windows”, the first full audio navigation for an operating system.

Christopher Reed

Silicon Valley veteran with proven ability to build and lead teams in both engineering and business organizations at companies ranging from startups to Fortune 50 companies.

Kieran Prior

Strategic and Investment expert with extensive business network. Goldman Sachs Proprietary Trading

Dr. Karoline Pershell

Dr. Pershell holds a PhD in Mathematics from Rice University. She is an expert in measuring and improving results through mathematical analysis.

get in touch