Our Artificial Intelligence: a technological framework fueling 7 core features
Applying AI to Human Resources (HR) is complex, and we had to develop our technology for over five years before it was able to address HR challenges.
Our career datasets contain over 58M possible job titles, including 80% that are unique. In order to understand all unique characteristics of a given career path it is necessary to have a firm understanding of semantic similarities, synonyms or homonyms in the job titles.
A challenge that only AI can fully address.
Skills-job correlation models
Clustree built Deep Learning networks able to match job titles to skills:
- Skills of other people with a similar job
- Skills developed during their work experience
Multiple skills detection models
We have 3 different models to detect skills in a document
- Search for skills in the textual content of the document
‘Data Science analysis in Python’ will result in the skills ‘Data Science’ and ‘Python’
- Search for possible complementary skills
Someone with ‘Talent Acquisition’ might have ‘Candidate Sourcing’ and ‘ATS’
- Search for skills implied by the job title and/or the organization
An employee having ‘Trader’ will also have ‘Derivatives’, ‘Hedging’ and ‘Portfolio management’
When detecting skills from a large referential, there is a high risk of false positives (wrongfully detected skills). We built some complex filter mechanisms in order to leverage a document’s context.
Elisa is a bench laboratory skill. It would be normal to detect it in the profile of a clinical trial associate. However it may also be the name of a coworker or a project on an accountant’s profile.
Proficiency level detection model
- Skills of other people with a similar job
We select the nearest skills using a skill distance
‘Sql Programming’ is a classic skill for a CTO, however, it is not a skill that is developed during this position - Skills developed during this experience
We factor in how recent the related positions are
Proficiency in a given skills tends to be lost if not practised for several years - And take into account the relative importance of skills for the position
Sales is more important for a ‘Head of Sales’ than for a ‘Product Manager’
Want to learn more about skills management?
Job transitions graph
Analyzing career paths is very valuable but it requires the understanding of thousands of job titles in an organization. Our AI pools similar jobs together (automatic clustering) at several granularity levels, in order to serve both local and global analysis:
- Global Transition Graphs
Vizualization of the most common career paths in the company
- Specific Population Graphs
Vizualization of career paths that lead to (or come from) a specific job family or corporate entity
Cuttingedge performance on skills search
- Standardized search: normalize a search query and find synonyms of similar skills
- Nearest skills: find skills that are statistically related to the skill being search for
- Job Title / Organizational skills: match a skill with a job title or an organization
- Fuzzy search: find results despite spelling errors
- Ranking by probability and expertise: show most likely results for the skill given the proficiency
Matching engine between jobs and profiles
Clustree computes matching scores from a combination of AI models based on skills & job titles, organizational specificities and employee wishes.
Matching engine for skills, learning and careers
- Outstanding skills required for a job: when an employee enters a job preference or looks at career opportunities, Clustree computes a set of missing skills to help in that transition.
- Most needed skills in the current job: when an employee seeks to develop her/his skills, Clustree suggests training in the skills that are most commonly developed when in that position.
- Most useful skills for career development: when an employee looks for career advice, Clustree predicts future positions and formulates a career plan with any training that would be necessary.
Over 400 000 HR and employees are using Clustree!