Connecting to LinkedIn...

Interview with Javid Muhammedali



Vice President, Artificial Intelligence


Bullhorn, the cloud computing company that helps staffing and recruiting organizations transform their businesses, recently hired Javid Muhammedali as its new Vice President, Product-Artificial Intelligence. Bullhorn has been training its machine learning and natural language processing models on the more than 600 billion records within the Bullhorn platform. Bullhorn is already the system of record for 10,000 customers around the world. Javid leads the company’s Artificial Intelligence (AI) business unit, a unit that operates as a well-funded and agile startup within the established and supportive Bullhorn environment.




How is AI changing the recruiting landscape?





Simon Childs (SC):  Javid can you tell us about what Bullhorn is doing with AI right now, and how that will help your extensive customer base?


Javid Muhammedali (JM): Let me explain first about the three core elements of AI we have at Bullhorn:


The first pillar is Pattern Recognition; how do you get the machine communicating with an actual word and extract the meaning of that word or the concept.  We are using it for job descriptions and more specifically the requirements needed to meet those job descriptions. Since Bullhorn has such a tremendous database of jobs, we have a lot of insight into what those job requirements entail. The AI we are building will help the recruiter ask much more nuanced and in-depth questions than for example, does the candidate have 3 years of java experience.


The middle pillar is Symbolic AI, something like a decision tree. At Bullhorn, we use Symbolic AI as a predictive model to go from observations about words used and then draw some inferences on how to better achieve the hiring outcome. This is configurable by users to tailor workflow and preferences to their needs.


The third pillar is Machine Learning, which is identifying connection points between the first two pillars, using very large-scale data analytics, and then applying that knowledge to create rules whereby the machine notices the correlation between them and automatically makes suggestions to the recruiter.


As for helping our customers, we are taking these 3 pillars and deploying them for specific uses:

For example, during the Screening Question Process, recognizing what the requirements are from a job description and then crafting through human-based training, good or optimal questions to ask a potential candidate, rather than presenting them with a generic candidate questionnaire.




SC: How does Cleo fit into all this?


JM: Cleo, which is something akin to “Alexa” or “Siri”, is Bullhorn’s own AI assistant deployed in the screening process. Having said that, we are expanding the horizon of what Cleo can do by making it a true Recruiter Assistant so that Cleo can help the recruiter interact with the system, help the candidate interact with the recruiter and help the recruiter interact with the hiring manager.  It is an Assistant to the Recruiter but not intended to be a replacement.


As I explained, our product strategy focuses on intelligence, workflow, and analytics, so with AI and with the analytics that comes with enough data from the screening questions, we can help benchmark recruiter productivity. How many candidates are the recruiters reaching out to, how many are they converting by the job, by the client, etc. and then feeding this data back into the system, so that when a new recruiter joins a company, they are not starting from scratch and they already have a base to work from.  The system also allows managers to more properly set KPI’s based on the intelligence gathered if that is required.


We are basing our machine learning algorithms on 3 things –

Firstly, who is the user and how are they behaving and performing over time.  Secondly, what is the job, for example, what is a project manager’s job, what are the skills required, how many candidates do you typically reach out to, how many do you interview and how many do you submit, etc.  The third thing is tying all the information gathered in the role back to the client, recognizing that the same title means different things in different organizations, and pinpointing the right sort of metrics for each client.






SC: When will all this new AI tech be available?


JM: We are shooting for the end of Q4 or early Q1, 2020. We are in the process of validating user cases with customers to make sure we have all the configuring done correctly.   We want to preserve what is unique about each of our clients and allow them, especially larger clients, to have the machine learning algorithms tuned towards the way they do business.


SC: Bullhorn has been quite aggressive in the acquisition space in recent years.  How are these newly acquired businesses adding value to your AI advancements?


JM: Let me give a couple of examples – we acquired Fyre which automatically captures candidates from third party systems and puts them into the ATS without the need for parsing. Also, which is geared towards executive search companies and in-house recruiting teams. Lots of cool AI came with these acquisitions, adding additional perspectives and value to our users.


Here :


Javid Muhammedali is an experienced senior executive with a track record of engineering excellence and business results. He has deep expertise in AI, natural language systems, unstructured data, and semantic search problems.  Over the course of his expansive career, he spent nine years at Monster, culminating in a role as VP of Product Management, during which he led search, partnerships and software solutions globally. More recently, Muhammedali was SVP of Integrated Solutions and Partnerships at, and a Knowledge Expert in Enterprise Software at McKinsey and Company.

Simon Childs currently acts as an Advisor to the Board for Recruit's International Recruitment Business and carries out additional strategic partnership, learning and development, and talent initiatives. He is an avid HR Tech investor/advisor to several start-ups and co-founded