Problem of matching people - talent with work opportunity - importance of a universal solution

People matching is one of the most exciting problems to solve. What can be more important to anyone than finding interesting work, or a job if one is unemployed, or a person with the right talent to do a task or shortlist for a job? Other subsets are matching benefactors and worthy causes or finding partners. Information & Communications Technology of today and a smart program can expand the search horizon manifold or find talent within a pool much more reliably and sensibly.

Automation and knowledge driven operations, changing life styles, rising aspirations and increasing mobility of people, apart from the enabling ICT are compelling motivators for building a smart people matching program and making it accessible for universal use. At Systems Dynamics, we have built such a program, that is TOTALLY INCLUSIVE (demographics agnostic) and SCALABLE. Called PEMS (People Expertise Matching System), it is configured and deployed for five models. Few use cases:

1.   Employee seeks interesting work opportunities within one’s organisation; Manager (or a co-worker) seeks the right talent (expert) to solve an urgent problem, work on a project or short assignment or fill a particular position.

2. Recruiter wants to hire candidate(s) to fill a new/vacant position.

3. Placement officer wants to place graduating students or find internships.

4. Empower Employment Facilitation Centres (EFC) to be located in Industrial Clusters where Work Seekers (WS) and Work Providers (WP) can walk in to register their profiles and requirements for receiving short-lists of matching opportunities or candidates.

5. Empower State or National level Employment Exchanges where WS and WP can register their profiles and requirements and expect to receive short-lists of matching opportunities or candidates (through SMS and email, as opted) and Government functionaries can obtain real time demand supply gap snapshots – region wise sector wise and receive data analytics that can inform their formulation of schemes, steer investments and resource to bridge the current and emerging gaps in the job market, upgrade skills, assist affirmative action groups or assist entrepreneurial ventures. In India today, ~ 1K Employment Exchanges (run by State Government’s Labour Departments) have ~ 40 million “unemployed WS” registrants but annually less than 1% get placed. With our solution, the successful placement rate can be easily quadrupled and registrants targeted to 400+ million “WS” within three years (“Work” is different from a “Job”). The social benefits are non-trivial enough, for justifying a national public sector project. Even a private venture could monetise some benefits and earn good ROI: The potential EVA (Economic Value Addition) could exceed 100 billion rupees p.a. (400 M WS registrants x 4% placement x average 6.5K monthly wage). The capital investment in empowering 1K employment exchanges and 100K Common Services Centres (existing in rural India), including training cost, will be less than 2 billion rupees! Operational cost will be mostly publicity and HelpDesk support. Low hanging fruit it is for Government or an Entrepreneur.

I will be happy to discuss the USP of PEMS (underlying system and method is patent pending) with people interested in evaluating its usage; or how to make PEMS universally accessible which is THE challenge – not if you happen to be Bill Gates or Larry Page or Anand Mahindra or Azim Premji ;) 

                                                                                                                                                    

                 

Note for Satya Nadella :)

LinkedIn has steadily improved in capturing human expertise and connecting people. Testimony is its growing popularity. However, it lacks the fine grained expertise mapping needed in today’s world to do automated matching. Enumerations in dynamic drop down lists, provided in current version of LinkedIn, do help the users escape the tyranny of “Key word matching” (as both WS and WP select from the same list) but more intelligence, greater precision and matching of attributes of preferences (other than expertise) are needed for sensible short-listing! It requires an ontology (parent-child relationship between expertise attributes) rather than linear drop down lists because the WS and WP may refer to skills in the same branch at different levels (for e.g.  SN’s project manager wants a developer in “.NET Framework” and the work seeker has mentioned her expertise as “LINQ” or let’s say, BG Foundation’s Manager wants to hire vaccination assistants and the work seeker has claimed experience in public health programs with a relevant healthcare certificate from National Council of Vocational Training).   

Note for Sundar Pichai ;)

Google’s mission is not just finding information on the web but information in the world! “People finding” is too alluring a problem to pass up for applying AI or Machine Learning which Google is brilliantly applying in its multiple products. But over here, this approach is NOT inclusive. In India 80% of workforce is not on web at all. Significant workers across the planet aren’t on the web. After Google and Google Earth it ought to be the turn of Google Mankind and not Google Web folks!

How to rein in terrorists using modern ICT

The trick is to build and deploy "loudly", systems which are visible deterrents for terrorists acts. The system described here can be implemented in any country because the solution is scalable.

Every resident in the country, and even a visitor to the country, should have a record created in a "REFERRALS DATABASE". This database can have millions or billion records (so even China and India are not excluded).

Referrals are people who have vouched for the subject (resident or the visitor). Each referee should provide a confidence score of the subject's probability of being or becoming a potential terrorist.

Each subject would be given a label in the database through an algorithm which would analyse the scores given by the referees and the missing data of the subject. The label instances would be:

Safe (not a likely terrorist)
Suspect (could be radicalised and become a terrorist)
Tag (is radicalised and a potential terrorist or has vital missing data and therefore in need of tagging)

Those labeled as "Tag" should be tagged (using active RFID non-removable bracelets) and subjected to tight surveillance, including geo fencing. Such people should be allowed to move only in designated areas or routes, and any deviation should alert security forces and trigger an arrest.

Those labeled "Suspect" should be asked to refresh referrals scores periodically (every 6 months or 2 years or some interval in-between depending upon the country's budget and propensity to rein in terrorists).


1. First the RESIDENTS' database:

1.1 The identifier in each RESIDENT'S record should be biometrics linked. So the Aadhar id qualifies in India. If the id does not have biometrics already linked, the country must collect that data of the subject (all fingerprints at least, both iris images desirable).

1.2 The subject's attributes should include - name, date of birth, place of birth, mobile number and email id (which have been authenticated with the identifier, for e.g. these should be consistent with the same data provided in Aadhar database - I am not suggesting stealing, or borrowing, this data from Aadhar database). A person who does not have either a mobile or email should be marked to "Tag".

1.3 REFEREE's scores - there should be 3 to 10 sets and each set of data should include identifier of the referee and his/her score on a 0 to 100 scale - 0 would indicate the referee's full confidence that the subject is not a potential terrorist at all, and a score of 100 would indicate the subject is radicalised, and is a terrorist, or a sure potential terrorist. There should be three to ten referees' scores, who are NOT themselves "Tag" type, depending upon the country's propensity to rein in terrorists and the budget allocated for building this system. The referee ids and scores would be initially provided by the subject himself or herself. The system would then authenticate those by obtaining confirmations from each referee by any convenient method - SMS, email or online.

2. The VISITOR'S database:

2.1 The identifier in each VISITOR's record will be Passport + Country issuing the passport.

2.2 Same as the data in item 1.2

2.3 Same as the data in item 1.3

The algorithm that delivers the verdict, Safe, Suspect or Tag, would evaluate the average scores of confidence by assigning weights to referees.

The above system would utilise advanced machine learning and communications technologies. Referees would be sent SMS and asked to confirm the scores input by the subject on their behalf.

Referees would be held responsible for patently wrong scores. Under a new legislation, a subject caught in an act of terrorism and assigned a score of less than 10 by any referee would automatically result in arraignment of the referee. Such a policy will be a deterrent to referees giving liberal scores. It would also act as a deterrent to would be terrorists, to indulge in terrorist acts which would bring harm to their friends and relatives. A terrorist is not afraid to lose his/her life but what about his/her friends' and relatives' lives?

The cost of the above system could be dropped by excluding such people who are considered "Safe". However, the drawback is that such categorisation attempts could get politicised easily. A centralised database of profiles of past terrorists could be built and analysed to determine common characteristics and any resident with matching characteristics would be required to register on the above REFERRALS DATABASE. For e.g. {all muslims or all those who have visited certain countries} AND {who are above 18 years but less than 80 years of age} could be considered "not safe" and ordered to register with REFERRALS DATABASE.