Engineering + Analytics = Product

E7d6e390a90513756419be75a43609ca?s=47 finid
June 29, 2019

Engineering + Analytics = Product

This presentation is a skit based on the graphic novels by Andrew Ng. In it, our heroes are caught between the worlds of analytics and engineering as they battle doing machine learning in the most inept way possible (i.e. – a business setting). Will our intrepid heroes actually pull it off? Find out and learn from our mistakes!

Key Takeaway Points:

1. Overcoming the language barrier
2. Testing business value
3. Knowing your teams strengths and weaknesses
4. Arrogance = Ego + Ignorance
5. Iteration helps

E7d6e390a90513756419be75a43609ca?s=128

finid

June 29, 2019
Tweet

Transcript

  1. None
  2. Big Data & AI Conference Dallas, Texas June 27 –

    29, 2019 www.BigDataAIconference.com
  3. Engineering + Analytics = Product Data Science Version

  4. Ron Dagdag – Sr. Software/Cloud Engineer at Crestron Electronics –

    Microsoft MVP award – Hackster DFW Ambassador meetup.com/Hackster-DFW – Dallas AR/VR Development meetup meetup.com/Dallas-Virtual-R eality Troy Mann - Risk Data Science Python Developer at Toyota Financial Services - Kaggle Junkie - Cloud Architect and Data Engineer - Meetup Junkie
  5. Disclaimer • All characters and other entities appearing in this

    work are fictittituts. Any resemblance to real persuns, living or dead, or other real-life entities, past or present, is purely coincidental. • Opinions expressed are solely our own and do not express the views or opinions of our employer.
  6. Agenda DATA COMPANIES AND MULTIDISCIPLINAR Y TEAMS EXPERTISE MISMATCH ENGINEER

    VS SCIENTIST DEFINITION OF “DONE”
  7. Stats For Normal Companies 75% of IoT projects in the

    US, UK, India are failing (Deloitte 2017) 85% of AI projects will fail (Gartner 2018) 60% of Big Data projects fail (Gartner 2016)
  8. Most Companies • Have No Strategy For: • Multidisciplinary teams

    • Data diversity • Correct Focus/No Magic Bullets • Tangible Business Problems Solved • Plan for User Adoption • Privacy and Security Plan
  9. Multidisciplinary Teams Include: PM and Management Data Science Functional (communicati

    on/ translation) Development/ Programming Data Engineering QA/Testing DevOps SME/Domain Expert
  10. Expertise Mismatch Does this sound familiar?

  11. Confict 1: Expertise Mismatch • Develuper Whu Wants tu Du

    Data Scitence Btt Duesn't Knuw Statitstitcs • Heuristic models (if/then) are AI too • Focused on Algorithms • No clue about statistical tests
  12. Confict 1: Expertise Mismatch Data Scitentitst Whu Stcks At Cuditng

    • Cannot write a function • No idea what objects are • Excel/Matlab/R Ninja
  13. It's OK Nobody knows everything Everyone must work together at

    their expertise
  14. Engineers vs Scientists Does this sound familiar?

  15. Confict 2: Engineers vs Scientists Engitneer/Develuper • Code • Test

    • Deploy Scitentitst • Measure • Test • Learn
  16. It's OK Engineers and Scientist have diferent ways of doing

    things and diferent goals. We need to respect that
  17. Defnition of “Done” Can you relate?

  18. Confict 3: Defnition of “Done” Prudtctitun cude its une thitng

    • Resilient • Scalable • Reliable • Repeatable Experitment cude its anuther • Correct problem • Statistically tested • Validated • Variables of high importance
  19. It's OK Sometimes we need to learn from our mistakes!

  20. Conclusion To be successful we need to think like a

    Data Company • The right resources to execute on valuable business projects • Understanding of AI enough to execute useful projects • Strategic direction to move to the future