How we scaled Geektrust to handle 100+ clients
I recently met a friend who’s worked at one of the largest HR consulting firms in the world. As we got talking, I mentioned we work with over 50 active companies with 0 recruitment associates, and he was taken aback.
A typical recruitment firm, he said, will have 1 recruitment associate to handle 5 clients and here we were handling 50 with 0 associates!
Later, reflecting on what we’ve been doing at Geektrust, I realized that we have no sales people, no dedicated marketing team, no recruitment associate, and still managed to scale to 100+ companies (and this is just the beginning).
Recruitment is typically a high touch business. You need to understand in-depth what each company does, what their requirements are, find candidates that match their needs, check profile quality, get them to solve coding challenges, evaluate the solutions, get companies to shortlist them, set up interviews, understand if an offer will be made, make sure the candidate joins once they accept the offer, and finally follow up to get payments on time.
As you can imagine, there are a lot of human touch points here. However, we felt the value of human interaction is 10x more for some touch points while we could apply tech to the remaining 90% of the processes.
Applying technology to recruitment
We didn’t start by applying tech to everything though. We started by doing things manually, understanding what works and then picking up problems to solve one by one.
When we started, we used to look at each profile to understand their core skills, their interests, how active they are, what sort of roles will fit them and which companies would be a match for them — we’ve automated this and now create metadata on the fly for the hundreds of developers who register with us weekly. We’ve used classifier systems (falling under the larger umbrella of machine learning) to implement this.
[Side note: — watch out for an article on how we did this. We have folks who fill in their profile saying they know java, android & testing. So what is their core skill? Mobile dev? Backend dev? or QA? Was an interesting problem to solve using ML
Edit: article is now published→ https://medium.com/@dhanushgopinath/how-we-used-machine-learning-to-identify-what-sort-of-technologists-register-on-geektrust-bdcbc59332db ].
We get about 500+ candidates expressing interest in companies monthly. How do you qualify them to make sure they’re a fit for each other? Both from a skill match as well as quality perspective? We used to do this manually but now we’ve automated this. Yet again, machine learning along with implementing a business rule engine helped us solve this.
Each company has its own nuances in processes and requirements. For example, it is not mandatory to have a resume to use Geektrust but for some companies it is mandatory. For some others they want all candidates to solve a coding challenge. Some others want to know your current salary. We used to go back and forth with candidates and companies on this. Now we’ve automated this. We’ve built a comprehensive business rule engine that is able to handle each company configuration and depending on the candidate metadata, automatically engages with the candidate appropriately.
From understanding candidates in-depth to qualifying applications to even asking for interview feedback, we’ve been finding opportunities and building a strong backend with extensive use of technology.
Our tech stack is Go, Python, Node, Elastic and AngularJS, and we use machine learning & recommender systems in our automation process.
Don’t start what you can’t automate
As a rule of thumb, we only do things that we know can be automated. For example — we don’t sign up for interviewing candidates even though many clients have asked us for that. Personal interviews cannot be automated, and therefore it’s not a part of what we offer our clients at Geektrust.
The next step for us is to automate our intensive code evaluation process. Code evaluation at the level we do has not been automated anywhere in the world. We are now using our learnings from the last 2 years of manual code evaluation to build our AI based code evaluation engine.
Interested in knowing more? Write to us at email@example.com
[About the author — Krishnan is co-founder at Geektrust.
Geektrust is built for technologists to connect with remarkable job opportunities.]