Cubra, the CV generator from Mindnow
Create a perfectly tailored CV for each application – professionally and at lightning speed. No problem with the help of ChatGPT. This is our exciting journey from zero to AI hero.
Article by
Jean-Paul Saija
How it all began
A cool CV and a crazy idea
One day, I needed a CV. I’m an entrepreneur and a creative free spirit – I had never actually created such a document before. A classic CV as a PDF was too boring for the occasion I needed it. How practical when you own a digital agency – our motivated team built me an entire website.
The CV website was great and went down really well with the target audience. This gave rise to the idea: Couldn't we develop a generator that anyone could use to create such a cool CV? I asked our UX Researcher to do a bit of research.
Section by Laura Antonietti
User research – indispensable in product development
Research of the recruitment workflow
During the research phase, we quickly realized that the idea of a CV website was too crazy. Recruiters have a certain workflow: They don't look at CVs that closely. They just screen for keywords because their job is stressful. Some jobs receive up to 180 applications and there is no time to look at each one individually.
PDFs are an absolute must
A CV is discarded if the keywords relevant to the position are not found in the first 5-7 seconds. An interactive website would make this workflow extremely difficult. Our conclusion: PDFs are boring, but an absolute must for applications.
A desire for better CVs
What also came out clearly in the research process was a desire for better CVs. Recruiters said that many applications were poorly made, with a confusing format and spelling mistakes. There was indeed a market for a CV tool.
What job seekers want
Our interviews were another proof of how important it is to conduct user research before developing a product. In the second step, we interviewed job seekers about how they created their CVs. It was people with and without a job, with lots and little experience, and from different industries. All these people had one thing in common: They didn't know where to start.
A tool for both sides
A bad CV is frustrating, both for the person who has difficulty creating it and for the person who has to evaluate the bad CV. Solving this problem was and is the basic idea behind Cubra.
But there were still a few technological hurdles to overcome before this could happen...
Section by Laura Antonietti
Section by Marko Martonosi
Senior Backend DeveloperHow does an AI-generated CV work?
The basis of letting Artificial Intelligence (AI) write your CV is algorithms that analyze data. The first step is therefore to provide data. On the one hand, this is the applicant's data, such as education and professional experience. On the other hand, there is a job advertisement. The AI analyzes both data sets and selects the most relevant keywords for the CV.
Algorithms also offer a major advantage for companies: They ensure that the person has the right qualifications, and is a “match”. This allows recruiters to find suitable candidates quickly.
Pulling data from LinkedIn
Algorithms work similarly for most CV generators out there. However, our Cubra software goes one step further: Instead of manually feeding the AI with personal data, all you need to do is copy the link from your LinkedIn profile.
With the help of ChatGPT, Cubra transfers all the information stored in the LinkedIn profile to the Cubra profile. No copying and pasting of documents, no synchronization of data and timelines, just a single link.
Working with AI requires problem-solving skills
An AI-based CV generator was new technological territory for our entire team. Despite this, or perhaps precisely because of this, everyone was immediately enthusiastic. Of course, we came up against a few hurdles.
Our journey from GPT-3 over GPT-3.5 to GPT-4
The technology on which Cubra is built has always been cutting-edge. When the development of Cubra started, GPT-3 was the latest AI. We needed data to finetune the way GPT-3 behaves and improve the quality of the generated CVs. We collected that data by allocating half of the Mindnow team to create as many CVs for as many different real job offers as possible.
When we were close to the result, GPT-3.5 came out and suddenly anyone could access trillions(!) of data through public APIs. This meant that the data we had painstakingly collected for the AI was suddenly completely superfluous. However, we didn't have time to be angry because we had to switch to the newest AI as quickly as possible.
The quality of prompted text improved
Currently, Cubra relies on GPT-4 – the latest OpenAI model and the most capable of all available AIs. It’s a bit slower than GPT-3.5, unfortunately. The delayed response is due to GPT-4 taking more time to compute a response. However, the result is considerable: No more gibberish, but high-quality text.
User-friendliness is the top priority
There has always been a lot of trial and error going on with Cubra. One idea was for users to be able to chat with Cubra. ChatGPT would ask questions so that users would no longer have to write anything themselves but simply click on an answer. But the user-friendliness could not keep up with the rapid technological development and this whole idea made Cubra more complicated to use and error-prone. So (for now) we switched back to the proven method of generating a CV in the first place and allowing the user to edit the document as they please.
Section by Marko Martonosi
Senior Backend DeveloperConclusion: Our curiosity and persistence taught us a lot
The fact that Cubra has ultimately become the tool that it is now is thanks to the unwavering curiosity of our great team, especially Marko. Our developers and designers went into the project with the motivation to reinvent the wheel and threw themselves into it. Time and again, we reached a point where the hurdles seemed insurmountable, but our team never tired of looking for solutions.
Cubra allows users to create an individual CV very quickly. And not just one, but one for each job you want to apply for. Highlighting the right keywords in each case increases the chances of a job interview many times over. More information on cubra.io
Our vision: Matching employers and job seekers
Besides overcoming technical obstacles, we learned that finding good people to hire is extremely difficult for companies. For certain positions, it takes months or even longer than a year.
Cubra is an online generator that anyone can use for free to create a professional CV. The advantage for recruiters is that they get to see better layouts with fewer typos.
For applicants, Cubra is and will remain free of charge. For companies, our vision is to provide a pool of skills, education, and experience that recruiters can access for a fee. Such an applicant pool could immediately provide recruiters with suitable talents without having to screen 180 CVs for keywords. With innovative AI tech, it’s possible. We’re working on it!
Mindnow AG
Weite Gasse 13
5400 Baden