Jobtome: Taking the job search to the next level with machine learning and AI
About Jobtome
Jobtome is an HR Tech company on a mission to help people get the right roles and employers find the right employees around the world.
Tell us your challenge. We're here to help.
Contact usAn integrated cloud platform allowed Jobtome to implement game-changing machine learning and AI quickly and simply, with minimum disruption and investment.
Google Cloud results
- AI-powered job recommendations that increased conversion rate by 5-10%
- X3 as many topics extracted using BERTopic algorithm and dynamic clustering
- Scalable system that could ingest 10 times the volume of data and produce 10 times as many recommendations
- Integrated solution that speeds up development time by 90%
Powering millions of smart live job recommendations
When searching for employment, it's critical that people have the tools to find a new job that matches their needs and skills quickly. Job search engine Jobtome realised that to differentiate itself in the job search market, and offer workers an effective service, it needed to provide matches that really worked. That meant identifying not just the keywords job seekers used, but signals from hundreds of other data points.
In order to understand and interpret all those signals to refine results, the technical team needed to consolidate their technical architecture into one managed platform, so they could better access metrics and data analytics. "We were trying to improve the way we were deploying software," says Paolo Santori, Chief Information and Data Officer, Jobtome. "So the capability to automate scale, manage software, get alerts, was critical."
"We were trying to improve the way we were deploying software. So the capability to automate scale, manage software, get alerts, was critical."
—Paolo Santori, CTO, JobtomeInitially, the Jobtome team was interested in the ability of the Google Kubernetes Engine (GKE) to manage containerized apps and help develop better and more reliable services for internal and external customers. But they quickly saw other potential benefits of Google Cloud.
"The data management capabilities of the platform were really the tipping point in our adoption [of Google Cloud]," says Santori. "That was the moment that we realised that it was giving us true power."
A suite of tools to build the infrastructure you need
The first thing the team at Jobtome set out to do was decouple the collection of analytical data from its applications. Up until that point, they'd been relying on operational data sets to extract metrics, but running a new stack on Google's Data Cloud meant they could leverage both the data warehousing power of BigQuery and Dataflow's ability to apply algorithms seamlessly, with no operational effort on their side.
"One of our jobs is to ingest, read, analyze, and categorize millions of jobs every day. The capability of doing it with a scaling platform like Dataflow was a game changer."
—Paolo Santori, CTO, JobtomeAs well as streaming and processing data, Dataflow also allowed Jobtome to scale its capabilities. "One of our jobs is to ingest, read, analyse, and categorise millions of jobs every day," says Santori. "The capability of doing it with a scaling platform like Dataflow was a game changer."
Adopting Vertex AI was the final step, as it meant Jobtome could apply its algorithms to a rich data set without the complexity of managing AI. That involves training a machine learning (ML) model, deploying it, storing and serving it, all of which takes time and effort, but with Vertex AI, they could leave all that to Google Cloud.
An integrated solution that makes decisions easy
Adding AI and ML capabilities was straightforward enough, as Vertex AI was already integrated with BigQuery and Dataflow. That freed up the team at Jobtome from having to find or build a solution that met 100% of its requirements, allowing them to focus on what they could do with it instead.
"We didn't have to rethink what we do," says Santori. "It was a natural evolution. This comes with an incredible value both on the technical side and the business side. Because of course, you don't have to invest a huge amount of money or a huge amount of time in a market where time and money are critical factors for success."
If anything, it was too easy, as Santori explains: "Sometimes developers can be a little bit scared of what they don't control end-to-end. So there was a little reluctance in adopting a solution that works like magic. 'How is it possible that if I plug in a model, I get the model served without doing anything?'"
As it turned out, it was just a matter of time before the developers at Jobtome came to trust the platform. They also had access to Google Cloud's customer engineering team, who gave them guidance and shared insights on a weekly basis, helping them identify faster and better solutions for their requirements.
An integrated solution also requires a level of trust, as Jobtome moved from assessing every option on the market each time it required new functionality, to simply choosing the relevant Google Cloud product instead. Santori believes it was this, as well as the close collaboration between the Jobtome and Google Cloud teams, that was key to the success of the project.
"This trust is changing the way we develop the applications," he says. "Now, if there is a solution in Google Cloud, our developers trust it, and they try to adopt it as a first choice. This is definitely making things much faster and easier."
Using machine learning and AI to create tangible business value
As a result, Jobtome had a stack that could gather, process, and serve terabytes of information about jobs, and scale ML to produce live, highly relevant recommendations. Santori estimates the conversion rate went up by 5-10%.
Implementing Google Cloud also meant Jobtome's system could deal with far greater volumes of data. Plugging in models served by Vertex AI meant it could ingest 10 times the number of jobs, from 1,000 to 10,000 jobs per second, while the number of recommendations it could produce went up from 10,000 to 100,000 per hour.
"We thought that our platform was optimised," says Santori. "But migrating it to Dataflow and Vertex AI made this optimization seem like a toy with respect to the new solution."
"More projects can go live, more tests, more experiments can be run, and more value can be extracted from the data that we have."
—Paolo Santori, CTO, JobtomeWorking with a modular system on Google Cloud also led to a reduction in the time Jobtome spent on deployment and integration, with data scientists or engineers spending 1/10 of the time it used to take to develop a new model's data set and make it available to the rest of the platform. And because all of these services are integrated, Jobtome can inject AI at any stage of the business or technical process - for example, the ETL process whereby data is extracted, transformed, and loaded into an output container.
The knock-on effect of all this is that Jobtome's time-to-market has become a fraction of what it used to be. More people in the company can extract value from the data and add valid ML contributions to the product, making it more democratic. "More projects can go live, more tests, more experiments can be run, and more value can be extracted from the data that we have," says Santori.
All in all, Jobtome's use of Google Cloud to build models, then create business value on top of those models, has transformed the way it does business, helping it move from being a data-driven company to one powered by the insights of AI.
Read more customer stories and use cases to find out how Google’s Data Cloud could help your business.
Tell us your challenge. We're here to help.
Contact us
About Jobtome
Jobtome is an HR Tech company on a mission to help people get the right roles and employers find the right employees around the world.