Employability Coach

Client Entrepreneur First

Contact:

Employers and investors will pay attention to your CV, but most are also going to take an interest in your online activities. The goal of this project is to create an employment/investment advisor that can process data from both CVs and online professional profiles, in order to estimate the potential value of different candidates. You will apply a range of natural language processing and machine learning techniques, to create a service that is valuable to anyone assessing a CV (their own, or others). The client will be able to supply an anonymised database of applicant data, and employment/investment decisions to correlate it with.

Initial discussions:

to make better and/or quicker selection decisions - has long been a goal for many employers. Advances in Natural Language Processing and other AI/ML techniques present an opportunity to make this a reality.

We invite you to explore whether hiring and selection can be improved using automated evaluation of written applications. You will explore different potential approaches to this problem and create prototype software that assesses and scores completed application forms. Interesting extensions might include exploring visualisation techniques to make the software's output transparent to non-technical users.

We see the software created having wide applicability and would encourage and support you to open source your work where appropriate.

From Alan, 5 September:

One thing that occurs to me is a privacy and data protection issue. Do you have permission from the people whose application data you are holding to use it in this way? Even if it has been submitted to you with consent for this kind of use, it might not be appropriate for personally identifiable data to be distributed to undergraduate students. We do have some provision for confidentiality of data in the group projects, but it might be necessary to either anonymise in some way, or for members of your organisation to be responsible for actual dataset handling.

The "spicing" ideas that I had were:

1. Make a comparison between the discriminatory power of application data and publicly available (Facebook) data. This in light of today's Dilbert! (link to follow)

2. If you're evaluating applicants primarily for entrepreneurship, then rather than simple yes/no, we could suggest that an enhanced algorithm offer a score for investment potential. (Of course, they'd have to notice that the judgement ground truth data they have is not necessarily well correlated with the value they need to derive)

3. Students might like to see something in it for them, rather than simply a sausage-selector for future recruiters. How about using the same scoring judgements to provide a CV-grooming service for job applicants?