How Wallbreakers Works

The Technology behind our upcoming platform

How Does Wallbreakers Work?

We are launching the wall breakers platform on April 30th. If you want to take a peak, email me and I will show. you.

To help major technology companies hire software engineers, we take on  candidates with cs degrees from 0-5 years of experience (mostly from underrepresented groups), and:

  • Use technology to analyze job openings

  • Use technology to categorize and score minority candidates’ hard and soft skills

  • Prepare them to do well in the hiring process

  • Present those candidates’ credentials to employers in a way that avoids them being subject to bias regarding their school, gender or race

Solution in Detail

Understanding job openings

We use natural language processing tools to extract keywords from job descriptions. For instance, in Exhibit 1, although the job description does not directly say so, we can deduce that this is a developer-evangelist role supporting multiple teams in a post-production environment. 

From that, we can deduce that the ideal candidate should have experience later in the systems development lifecycle (SDLC), such as in a release-engineering role. They must also have the communication skills to receive and respond to feedback from multiple customers. The requirement to quickly become familiar with the company’s APIs will likely require  them to be fluent in Javascript, and Java and Python would be nice-to-have skills. Other nice-to-haves we can infer include experience in testing, network programming, and technical writing.

Exhibit 1

We use machine learning (ML) capable of extracting the smallest units of meaning from language through the context in which they appear.  Hard and soft skills needed for the jobs are revealed by inference from this linguistic breakdown, allowing us to predict which skill sets are most likely to be successful in both the interview process and the job setting.

Ranking candidates’ skills

Also using NLP, we extract candidates’ hard skills from their resumes. We derive insights about a candidate’s soft skills and likely fit in a team from NLP analysis of group interactions and mock interviews. For instance, we look for keywords associated with soft skills such as leadership and team-building, and confidence. We quantify these for candidates as in Exhibit 2.

Exhibit 2

Correlating characteristics that matter

Then we compare the candidates’ hard and soft skills to the job requirements to determine the best fits and present those to employers.

We do not compare candidates with a set of previously “successful” candidates. We only compare them with the other current candidates, so there’s no danger of our algorithms perpetuating an existing biased process.

Exhibit 3

Training to fill the gaps

We give every candidate six weeks of training before they go to interviews. New computer science graduates, especially from lower-profile schools but even from highly selective schools, often are not well-prepared for the interview process. They frequently lack experience in hands-on coding and real-time problem-solving, both of which are commonly tested. So, the training includes individual and group problem solving; hands-on coding exercises; technical topics such as space-and-time complexity analysis, networking, Linux, and multithreaded programming; and one-on-one mock interviews with senior software engineers.  

For more senior engineers, we only use GitHub or other evidence of performance instead of the training.

Results

80% of candidates we refer are given the first interview. We average 68% referral to offer ratio. That compares with an average of 10% for minority applicants and 20 percent for general applicants.