Initiative for Analytics and Data Science Standards (IADSS) workshop presentation at the ACM KDD conference (Association of Computing Machinery Knowledge Discovery in Databases).
Kdd 2019: Standardizing Data Science to Help Hiring
1. STANDARDIZING DATA SCIENCE TO HELP HIRING,
BY GREG MAKOWSKI
ACM KDD Conference 2019
Initiative for Analytics and Data Science Standards (IADSS) Workshop
https://www.iadss.org/sigkdd-2019
August 5, 2019
2. Problem:
• This talk discusses the desires of a hiring manager for standards in the Data
Science profession – to help with growing teams
Outline:
• Challenges and motivation
• Way to Standardize a Profession – Testing
• Way to Share Experience – a Portfolio
• How can we have a “Searchable Portfolio”?
• Is a privacy preserving component helpful? (i.e. references to confirm)
2
3. Pain Points in DS Hiring
• To hire people to work on DS consulting projects, it is good to find people
with X (i.e. 3+) past DS projects DEPLOYED
• It is very hard to get good support to find qualified candidates.
(to search for complex past experiences)
• HR
• Recruiters
3
4. Way to Standardize a Profession – Testing
• How do other professions “standardize” to a consistent experience level?
• Accounting – Certified Professional Accountant (CPA)
• Medical Dr. - United States Medical Licensure Exam (USMLE)
• For DS, testing may be a better option for those just out of school
• Do we form a group to decide testing standards?
• ACM KDD Data Science Exam (KDD DSE)?
• All DS pass 5 core skills?
• Optionally pass 0+ of 15 optional skills (like medical specialties?)
• For DS, is this most helpful before the person has deployed a few projects?
4
5. Way to Share Experience – a Portfolio
• Other professions have other “standards for experience” besides testing
• Medical doctors go through a residency – a mentored practical training program
• Medical doctors have to make life and death decisions
• It may not be a quick adjustment to REQUIRE Data Scientists to go through.
a residency
• Other professions that have portfolios…
• Research, teaching, architecture, photography, …
• Informally, DS’s have this in their LinkedIn profile
• Challenge – you can’t search for
• Candidates who have deployed at least 3 data mining models
5
6. How can we have a “Searchable Portfolio”?
• Can we go from unstructured text to structured JSON key-values?
• Github examples? https://paperswithcode.com/ ? What is original work?
• Problem
• Title
• Context
• Vertical market
• Problem type
• Solution
• Algorithm family
• Algorithm(s)
• Key innovations
• Result
• Intended to be deployed?
• Was deployed
• Estimated annual value?
6
7. Is a privacy preserving component helpful?
• Currently, some confidential sharing can happen
• Past managers provide references
• Clients that don’t agree yet to press releases can be sales references by phone
• GENERALIZATION: Some information is not openly shared with the public,
but under the right circumstances, privately shared.
• Is there a way to use crypto-sharing (i.e. bitcoin style), such as between
hiring managers?
• Between a past client and a hiring manager or future prospective client?
• Who manages the structured portfolio database? (for more than DS?)
7
8. Can we create structured portfolios?
• Discussion
• Does it sound useful to be able to query portfolios and skills?
• What are ways to standardize DS as a profession?
www.LinkedIn.com/in/GregMakowski
8