The Secret Sauce of Successful
Data Science Teams
Ganes Kesari Webinar, Jan 2020Sundeep Reddy
2
WITH A JOB DESCRIPTION LIKE THIS..
- Data collection experience
- Data processing, cleaning, and transformation
- Features engineering
- Programming skills, DevOps & deployment
- Expert in Statistical analysis
- Excellent Machine learning skills
- Hands-on in Deep learning
- Strong in Data visualization
- Problem-solving skills
- Strong domain expertise & business analysis
skills
- Reporting and good communication
- Ability to manage and mentor a small team
…you’re really looking for
A data Science
Department!
Not a “Full Stack data
Scientist”
3
FEELING LUCKY? HERE’S A DATA SCIENCE TITLE GENERATOR!
Data
Statistical
ML
AI
Chief
Principal
Senior
Junior
Associate
Deputy
Assistant
Scientist
Engineer
Analyst
Designer
Developer
Designer
Storyteller
Ninja
Chef
Wrangler
Evangelist
Rock Star
Wizard
Alchemist
Senior Data ScientistPrincipal AI StorytellerChief Data Wizard
4
BUZZWORDS AND BUSTED BUDGETS
5
ROLES &
SKILLS
EMERGING
ROLES
HIRING
TIPS
6
INTRODUCTION
Ganes Kesari
Co-founder & Head of Analytics
100+ ClientsInsights as Stories
Help start, apply and adopt Data Analytics &
Visualization for Business impact
Sundeep Reddy
SVP, Hiring & Products
@kesaritweets
/gkesari
@sundeeprm
/sundeepreddym
7
POLL #1
DATA SCIENCE ROLES
Which of the following roles exist in your data science teams?
(Choose all that apply)
➢Domain experts fluent in data
➢Experts in Statistics and machine learning
➢Programmers with front-end / back-end skills
➢Programmers who can productionize ML models
➢Experts in Information design / UX / Visual design
8
MATURITY LEVELS WITH DATA
Data Engineering
ActivitiesMaturityPhases
Data Science
Data as
‘Culture’
Data
Collection
Data
Storage
Data
Transformatio
n
Reporting Insights Consumption Decisions
Logs, IOT
Int/External
Stage/Strea
m
SQL, Spark..
Un/Structure
d
Data lake..
Cleaning
ETL
Preparation
Aggregates
Metrics/KP
I
Reports
ML
EDA
AI
Info Design
Narrative
Data Stories
Workflows
Change Mgmt
Actions
https://techcrunch.com/2019/12/13/when-and-how-to-build-out-your-data-science-team/
9
MATURITY LEVELS WITH DATA
Data Engineering Data Science
Data as
‘Culture’
Data
Collection
Data
Storage
Data
Transformatio
n
Reporting Insights Consumption Decisions
Logs, IOT
Int/External
Stage/Strea
m
SQL, Spark..
Un/Structure
d
Data lake..
Cleaning
ETL
Preparation
Aggregates
Metrics/KP
I
Reports
ML
EDA
AI
Info Design
Narrative
Data Stories
Workflows
Change Mgmt
Actions
ActivitiesMaturityPhases
https://techcrunch.com/2019/12/13/when-and-how-to-build-out-your-data-science-team/
10
DATA SCIENCE SOLUTION: EXAMPLE
https://gramener.com/cargo/
11
AND HERE ARE THE BUILDING BLOCKS OF THIS EXAMPLE
Domain
Design
Analytics
Development
• Impact analytics
• What-if modelling
• Business workflow
• Influencing factors
• Data transformation
• Frontend/backend coding
• User journey
• Visuals & aesthetics
12
FIVE ROLES
& SKILLS IN
DATA SCIENCE
1
13
PRE-REQUISITE FOR EVERY ROLE IN DATA SCIENCE
Data
Science
Literacy
Passion
for data
Inferences
with data
Domain
awareness
Eye for
design
https://towardsdatascience.com/whats-the-secret-sauce-to-transforming-into-a-unicorn-in-data-science-94082b01c39d
14
5 ROLES & SKILLS IN DATA SCIENCE
1. Data Translator 2. Data Scientist
- Own from inception to adoption
- Translate across domain & data
- Act as a glue in the team
- Domain expertise
- Business analysis & solutioning
- Interpersonal & mentoring skills
Business analyst, Domain experts
- Devise analytics approach
- Analyze data & identify insights
- Build ML models
- Statistics and machine learning
- Identify & interpret insights
- Scripting skills
Statistician, ML experts
Responsibilitie
s
Skills
Closest role
https://techhq.com/2019/12/a-complete-data-science-team-requires-more-than-just-data-scientists/
15
5 ROLES & SKILLS IN DATA SCIENCE
- Ensure consumption of insights
- Design information architecture
- Understand user, drive adoption
- Information design
- User centered design
- Aspects of interface/visual design
UX Designer, Interaction designer
- Package data science solution
- Productionizing, DevOps
- Data pipelines/integration
- Software engineering
- Data handling
- Front-end / Back-end coding
Software engineer, Data architect
Responsibilitie
s
Skills
Closest role
3. Information Designer 4. ML Engineer
https://techhq.com/2019/12/a-complete-data-science-team-requires-more-than-just-data-scientists/
16
5 ROLES & SKILLS IN DATA SCIENCE
- Identify roadmap & scale maturity
- Ensure biz value from data science
- Drive a culture of data
- Project management
- Business analysis, solutioning
- Team handling
Project manager, Business analyst
Responsibilitie
s
Skills
Closest role
5. Data Science Manager
https://techhq.com/2019/12/a-complete-data-science-team-requires-more-than-just-data-scientists/
17
HOW TO CREATE AN ENVIRONMENT FOR TEAM SUCCESS?
• Promote collaboration within teams
• Align closely with business users
• Measure outcomes through business value
• Adopt process frameworks for consistency
• Up-skill continuously on tech skillsets
• Nudge cross-training across disciplines
18
5 CORE ROLES IN DATA SCIENCE
Data
Translator
ML Engineer
Information
Designer
Data Scientist
Data Science Manager
19
POLL #2
EMERGING ROLES IN DATA SCIENCE
In your opinion, which of these non-core disciplines can add
most value to data science? (Choose all that apply)
➢Creative arts
➢Social science
➢Journalism
➢Law
➢Humanities
20
THREE
EMERGING
ROLES
2
21
1. DATA STORYTELLER
Role Highlights
- Dashboards are NOT data stories
- Stories=visual+context+narrative
- Fields: Journalism, creative arts
https://gramener.com/playground/senate/similarity
22
2. BEHAVIORAL PSYCHOLOGIST
Role Highlights
- Human side of data insights
- More practical, ‘accurate’ results
- Fields: Social sciences
Gramener Telecom case study
23
3. DATA ETHICIST
https://gramener.com/emailnetwork/
Role Highlights
- Ensure trust & fairness
- Act as a collective conscience
- Fields: Law, Humanities
This visual shows the network of email
exchanges between people.
Look for the closest neighbors. The
distance is a function of email exchange.
A top BFSI player wanted a scientific way
to identify peers, for employee feedback.
Was there an alternative to manually
screening for peer review?
24
3 EMERGING SKILLS IN DATA SCIENCE
https://towardsdatascience.com/the-3-missing-roles-that-every-data-science-team-needs-to-hire-97154cc6c365
1. Data Storyteller 2. Behavioral Psychologist
3. Data Ethicist
25
RECAP: 8 DATA SCIENCE ROLES ACROSS MATURITY LEVELS
Data Engineering
ActivitiesMaturityPhases
Logs, IOT
Int/External
SQL, Spark..
Un/Structure
d
Cleaning
ETL
Aggregates
Metrics/KP
I
ML
EDA
Data Stories
Info Design
Data Science
Data as
‘Culture’
Data
Collection
Data Storage
Data
Transformatio
n
Reporting Insights Consumption Decisions
Stage/Stream Data lake.. Preparation Reports AI Packaged App
Actions
Workflows
Change Mgmt
Data Scientist
Data Translator
Data Science
Info Designer
Storyteller
ML Engineer
Behavioral
Psychologist
Data Ethicist
Gramener
26
HIRING TIPS &
TRICKS
3
27
THE STATE OF DATA SCIENCE/ ML / AI TALENT GAP
We will NOT fill 100% of Talent Demand
28
THE DEMAND PICTURE
Source: Linkedin.com
Job Posting: Data Science
Period: Last one month
Location: USA
Total: 55,469
1. Type of Job 2. Work Location
3. Skill Level 4. Who is hiring
29
THE SUPPLY FUNNEL
Where can we find this talent What kind of companies do they work for?
* Hacker Rank – 2019 Developers Skills Report * Kaggle – 2019 DS & ML Survey
30
COMMON RECRUITMENT CHALLENGES
1. Padded Resumes
2. Salary Expectations
3. Job Hoppers
4. Geography
5. Industry
6. Lack of Soft Skills
7. Lack of Domain Skills
8. You can’t justify Full-time hire
31
HOW TO OFFSET TALENT SHORTAGE
HOW TO SPEED UP TIME TO HIRE
HOW TO ACCELERATE BUSINESS TRANSFORMATION
32
WHERE DO WE SCOUT FOR TALENT
Careers Page, Employee Referrals, Job Portals
Regional Meetups & Networking events
Conference – NIPS, ACM, IEEE etc.
Consultants and Freelancers
Partnership with Colleges &
Universities
- Interns for specific duration
- Steady supply of Entry level talent
- Long term Partnership
- Short term Consulting engagements
Internal Talent
- Members with near by skills
- Members willing to change roles
- Providing training and opportunity
33
SCREENING PROCESS BY ROLE
Data
Translator
ML Engineer
Information
Designer
Data Scientist
Data Science Manager
34
WHAT CAN WE DO DURING INTERVIEW PROCESS
• Provide clarity on the Role and Job location
• Be explicit about projects they will work on, Career progression, flexibility offered
• Explain the interview process and expectation at each step
• Asking irrelevant questions
• Doing no or slow follow ups
• Set expectation of monetary compensation offered for the role
35
PATH AFTER GETTING ONBOARD
• Care for Onboarding team members
• Have a structured Orientation Plan to cover – Organization, Business, Function and
Team goals
• Offer the team Challenging/ Complex problems to work on
• Upskill team skills across
• Communication & Intern-personal
• Business Acumen
• Software Engineering skills
• Provide opportunity for networking with Stakeholders
36
TO LEARN MORE.. GRAMENER WEBINARS
https://register.gotowebinar.com/recording/4757207884753493515
The Best way to Choose your
Data Science Projects
How to Organize data science Teams
for Effective Collaboration
Completed.
Recording available!
Watch out for the dates!
37
Thank You!
@kesaritweets
/gkesari
@sundeeprm
/sundeepreddym
Get in touch for help with your data science
projects or to build your team

How to Build Data Science Teams

  • 1.
    The Secret Sauceof Successful Data Science Teams Ganes Kesari Webinar, Jan 2020Sundeep Reddy
  • 2.
    2 WITH A JOBDESCRIPTION LIKE THIS.. - Data collection experience - Data processing, cleaning, and transformation - Features engineering - Programming skills, DevOps & deployment - Expert in Statistical analysis - Excellent Machine learning skills - Hands-on in Deep learning - Strong in Data visualization - Problem-solving skills - Strong domain expertise & business analysis skills - Reporting and good communication - Ability to manage and mentor a small team …you’re really looking for A data Science Department! Not a “Full Stack data Scientist”
  • 3.
    3 FEELING LUCKY? HERE’SA DATA SCIENCE TITLE GENERATOR! Data Statistical ML AI Chief Principal Senior Junior Associate Deputy Assistant Scientist Engineer Analyst Designer Developer Designer Storyteller Ninja Chef Wrangler Evangelist Rock Star Wizard Alchemist Senior Data ScientistPrincipal AI StorytellerChief Data Wizard
  • 4.
  • 5.
  • 6.
    6 INTRODUCTION Ganes Kesari Co-founder &Head of Analytics 100+ ClientsInsights as Stories Help start, apply and adopt Data Analytics & Visualization for Business impact Sundeep Reddy SVP, Hiring & Products @kesaritweets /gkesari @sundeeprm /sundeepreddym
  • 7.
    7 POLL #1 DATA SCIENCEROLES Which of the following roles exist in your data science teams? (Choose all that apply) ➢Domain experts fluent in data ➢Experts in Statistics and machine learning ➢Programmers with front-end / back-end skills ➢Programmers who can productionize ML models ➢Experts in Information design / UX / Visual design
  • 8.
    8 MATURITY LEVELS WITHDATA Data Engineering ActivitiesMaturityPhases Data Science Data as ‘Culture’ Data Collection Data Storage Data Transformatio n Reporting Insights Consumption Decisions Logs, IOT Int/External Stage/Strea m SQL, Spark.. Un/Structure d Data lake.. Cleaning ETL Preparation Aggregates Metrics/KP I Reports ML EDA AI Info Design Narrative Data Stories Workflows Change Mgmt Actions https://techcrunch.com/2019/12/13/when-and-how-to-build-out-your-data-science-team/
  • 9.
    9 MATURITY LEVELS WITHDATA Data Engineering Data Science Data as ‘Culture’ Data Collection Data Storage Data Transformatio n Reporting Insights Consumption Decisions Logs, IOT Int/External Stage/Strea m SQL, Spark.. Un/Structure d Data lake.. Cleaning ETL Preparation Aggregates Metrics/KP I Reports ML EDA AI Info Design Narrative Data Stories Workflows Change Mgmt Actions ActivitiesMaturityPhases https://techcrunch.com/2019/12/13/when-and-how-to-build-out-your-data-science-team/
  • 10.
    10 DATA SCIENCE SOLUTION:EXAMPLE https://gramener.com/cargo/
  • 11.
    11 AND HERE ARETHE BUILDING BLOCKS OF THIS EXAMPLE Domain Design Analytics Development • Impact analytics • What-if modelling • Business workflow • Influencing factors • Data transformation • Frontend/backend coding • User journey • Visuals & aesthetics
  • 12.
    12 FIVE ROLES & SKILLSIN DATA SCIENCE 1
  • 13.
    13 PRE-REQUISITE FOR EVERYROLE IN DATA SCIENCE Data Science Literacy Passion for data Inferences with data Domain awareness Eye for design https://towardsdatascience.com/whats-the-secret-sauce-to-transforming-into-a-unicorn-in-data-science-94082b01c39d
  • 14.
    14 5 ROLES &SKILLS IN DATA SCIENCE 1. Data Translator 2. Data Scientist - Own from inception to adoption - Translate across domain & data - Act as a glue in the team - Domain expertise - Business analysis & solutioning - Interpersonal & mentoring skills Business analyst, Domain experts - Devise analytics approach - Analyze data & identify insights - Build ML models - Statistics and machine learning - Identify & interpret insights - Scripting skills Statistician, ML experts Responsibilitie s Skills Closest role https://techhq.com/2019/12/a-complete-data-science-team-requires-more-than-just-data-scientists/
  • 15.
    15 5 ROLES &SKILLS IN DATA SCIENCE - Ensure consumption of insights - Design information architecture - Understand user, drive adoption - Information design - User centered design - Aspects of interface/visual design UX Designer, Interaction designer - Package data science solution - Productionizing, DevOps - Data pipelines/integration - Software engineering - Data handling - Front-end / Back-end coding Software engineer, Data architect Responsibilitie s Skills Closest role 3. Information Designer 4. ML Engineer https://techhq.com/2019/12/a-complete-data-science-team-requires-more-than-just-data-scientists/
  • 16.
    16 5 ROLES &SKILLS IN DATA SCIENCE - Identify roadmap & scale maturity - Ensure biz value from data science - Drive a culture of data - Project management - Business analysis, solutioning - Team handling Project manager, Business analyst Responsibilitie s Skills Closest role 5. Data Science Manager https://techhq.com/2019/12/a-complete-data-science-team-requires-more-than-just-data-scientists/
  • 17.
    17 HOW TO CREATEAN ENVIRONMENT FOR TEAM SUCCESS? • Promote collaboration within teams • Align closely with business users • Measure outcomes through business value • Adopt process frameworks for consistency • Up-skill continuously on tech skillsets • Nudge cross-training across disciplines
  • 18.
    18 5 CORE ROLESIN DATA SCIENCE Data Translator ML Engineer Information Designer Data Scientist Data Science Manager
  • 19.
    19 POLL #2 EMERGING ROLESIN DATA SCIENCE In your opinion, which of these non-core disciplines can add most value to data science? (Choose all that apply) ➢Creative arts ➢Social science ➢Journalism ➢Law ➢Humanities
  • 20.
  • 21.
    21 1. DATA STORYTELLER RoleHighlights - Dashboards are NOT data stories - Stories=visual+context+narrative - Fields: Journalism, creative arts https://gramener.com/playground/senate/similarity
  • 22.
    22 2. BEHAVIORAL PSYCHOLOGIST RoleHighlights - Human side of data insights - More practical, ‘accurate’ results - Fields: Social sciences Gramener Telecom case study
  • 23.
    23 3. DATA ETHICIST https://gramener.com/emailnetwork/ RoleHighlights - Ensure trust & fairness - Act as a collective conscience - Fields: Law, Humanities This visual shows the network of email exchanges between people. Look for the closest neighbors. The distance is a function of email exchange. A top BFSI player wanted a scientific way to identify peers, for employee feedback. Was there an alternative to manually screening for peer review?
  • 24.
    24 3 EMERGING SKILLSIN DATA SCIENCE https://towardsdatascience.com/the-3-missing-roles-that-every-data-science-team-needs-to-hire-97154cc6c365 1. Data Storyteller 2. Behavioral Psychologist 3. Data Ethicist
  • 25.
    25 RECAP: 8 DATASCIENCE ROLES ACROSS MATURITY LEVELS Data Engineering ActivitiesMaturityPhases Logs, IOT Int/External SQL, Spark.. Un/Structure d Cleaning ETL Aggregates Metrics/KP I ML EDA Data Stories Info Design Data Science Data as ‘Culture’ Data Collection Data Storage Data Transformatio n Reporting Insights Consumption Decisions Stage/Stream Data lake.. Preparation Reports AI Packaged App Actions Workflows Change Mgmt Data Scientist Data Translator Data Science Info Designer Storyteller ML Engineer Behavioral Psychologist Data Ethicist Gramener
  • 26.
  • 27.
    27 THE STATE OFDATA SCIENCE/ ML / AI TALENT GAP We will NOT fill 100% of Talent Demand
  • 28.
    28 THE DEMAND PICTURE Source:Linkedin.com Job Posting: Data Science Period: Last one month Location: USA Total: 55,469 1. Type of Job 2. Work Location 3. Skill Level 4. Who is hiring
  • 29.
    29 THE SUPPLY FUNNEL Wherecan we find this talent What kind of companies do they work for? * Hacker Rank – 2019 Developers Skills Report * Kaggle – 2019 DS & ML Survey
  • 30.
    30 COMMON RECRUITMENT CHALLENGES 1.Padded Resumes 2. Salary Expectations 3. Job Hoppers 4. Geography 5. Industry 6. Lack of Soft Skills 7. Lack of Domain Skills 8. You can’t justify Full-time hire
  • 31.
    31 HOW TO OFFSETTALENT SHORTAGE HOW TO SPEED UP TIME TO HIRE HOW TO ACCELERATE BUSINESS TRANSFORMATION
  • 32.
    32 WHERE DO WESCOUT FOR TALENT Careers Page, Employee Referrals, Job Portals Regional Meetups & Networking events Conference – NIPS, ACM, IEEE etc. Consultants and Freelancers Partnership with Colleges & Universities - Interns for specific duration - Steady supply of Entry level talent - Long term Partnership - Short term Consulting engagements Internal Talent - Members with near by skills - Members willing to change roles - Providing training and opportunity
  • 33.
    33 SCREENING PROCESS BYROLE Data Translator ML Engineer Information Designer Data Scientist Data Science Manager
  • 34.
    34 WHAT CAN WEDO DURING INTERVIEW PROCESS • Provide clarity on the Role and Job location • Be explicit about projects they will work on, Career progression, flexibility offered • Explain the interview process and expectation at each step • Asking irrelevant questions • Doing no or slow follow ups • Set expectation of monetary compensation offered for the role
  • 35.
    35 PATH AFTER GETTINGONBOARD • Care for Onboarding team members • Have a structured Orientation Plan to cover – Organization, Business, Function and Team goals • Offer the team Challenging/ Complex problems to work on • Upskill team skills across • Communication & Intern-personal • Business Acumen • Software Engineering skills • Provide opportunity for networking with Stakeholders
  • 36.
    36 TO LEARN MORE..GRAMENER WEBINARS https://register.gotowebinar.com/recording/4757207884753493515 The Best way to Choose your Data Science Projects How to Organize data science Teams for Effective Collaboration Completed. Recording available! Watch out for the dates!
  • 37.
    37 Thank You! @kesaritweets /gkesari @sundeeprm /sundeepreddym Get intouch for help with your data science projects or to build your team