Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.
Upcoming SlideShare
What to Upload to SlideShare
What to Upload to SlideShare
Loading in …3
×
1 of 24

Tips for Effective Data Science in the Enterprise

0

Share

Data Science is an evolving field, that requires a diverse skill set. From Career Advice to steps for how to approach your Data Science Workflow, this talk is full of practical tips that you can apply immediately to your job.

Tips for Effective Data Science in the Enterprise

  1. 1. Tips for Effective Data Science in the Enterprise Lisa Cohen
  2. 2. 2 Session goals • Demystify Data Science Career Paths • Discuss best practices to tackle a Data Science Project • Gain Tips & Tricks for DS scenarios in the enterprise
  3. 3. 2000 Applied Math Bachelor & Masters degrees Quantitative thinking, Applied sciences Cum laude 2004 VS Languages & IDE Technical Feature PM Building software, understanding customers, leading cross-functional feature team Incorporating SQL & data access into .NET programming languages Shipped VS 2005, 2008, 2010 Reached ~9M customers 3 patents for new designs 2012 Sr Mgr – VS Telemetry Product analytics: working with perf, compatibility, planning, privacy, compliance Led DevDiv business reviews, Advanced VS telemetry, Delivered VS active use clustering, Launched “Send A Smile” feedback 2015 Principal Data Scientist Mgr Using data to help customers succeed & grow on Azure. Driving DS best practices, Cross-MS partnerships Led cross-functional team. Evolved DS for credit offers, partners, Direct customers, Support. 2008 Sr Community PM - VS Divisional strategy, industry trends, Central org & systems, Cross group partners, Exec comms, CSAT & customer/ partner communities Led DevDiv blogs (3M+ views/mo), Presented keynotes & sessions at 50+ conferences YR Role Experience& Learnings Achievements
  4. 4. Nurturing customers, growing the business, connecting sources and advancing the DS craft Students & Developers Student offer, VSE Direct / Unmanaged Individuals, SMC, Sign-Up Field & EA Enable Sales Cross Cloud, Support, Retention, Service Usage, Fraud, Payments Data Science, Machine Learning, ML Ops, Experimentation, Data Vis, PM Partners & Startups ISVs, CSP, PAL, DPOR, MfS Marketplace App Source Partner Center Customer Portal Azure Portal Advisor Cost Management Azure.com Docs Learn PORTALSAUDIENCES FUNDA- MENTALS CAP- ABILITY
  5. 5. 6 Session goals • Demystify Data Science Career Paths • Discuss best practices to tackle a Data Science Project • Gain Tips & Tricks for DS scenarios in the enterprise
  6. 6. Demystifying Career Paths  Roles: What excites you about Data Science? Data Scientist Analytics & Inference • Statistical analysis & experiments Machine Learning Scientist/Engineer Production Models • Develop predictive models, MLOps Data Engineer Data Platform & Pipelines • Build the data platform Program Manager Planning & Stakeholder Engagement • Manage the data science process Tip #1: Follow your passion
  7. 7. The Data Science Venn Diagram  Technical  Analytical Problem Solving  Statistics  Querying  R, Python, SQL, Kusto  Big Data  Modeling  Data Visualization Technical Soft Skills Domain  Domain  Business context  Data sets  Soft Skills  Communication  Organization  Cross-Group Collaboration  Teamwork Conway’s Venn Diagram Tip #2: Chart your path
  8. 8. Data Science Organizations  What kind of environment do you want to be in? CentralizedEmbedded A core data science org provides services to business or functional teams across the company as a center of excellence Individual data science teams are spread throughout the company, reporting to and serving specific business or functional teams Tip #3: Find a DS community Tip #4: Connect with Stakeholders
  9. 9. 10 Project Intake Tips Kicking off a model, experiment or analytics project  What new capability will this enable?  What decision/action will you take?  What’s the expected impact? Planning Process Project Intake Questions Prioritization & Scalable Solutions Tip #5: Focus on what matters (Prioritize with stakeholders, ask questions, socialize results)
  10. 10. Data Science Lifecycle Problem & Hypothesis Design Approach Data Acquisition & Exploration Analysis & Predictive Modeling Evaluation & Reviews Deployment & Socialization Data Science Lifecycle MS Doc: Team Data Science Process
  11. 11. Data Science Lifecycle Problem & Hypothesis Design Approach Data Acquisition & Exploration Analysis & Predictive Modeling Evaluation & Reviews Deployment & Socialization Data Science Lifecycle MS Doc: Team Data Science Process
  12. 12. Data Science Lifecycle Problem & Hypothesis Design Approach Data Acquisition & Exploration Analysis & Predictive Modeling Evaluation & Reviews Deployment & Socialization Data Science Lifecycle MS Doc: Team Data Science Process
  13. 13. Explore the underlying data  Explore completeness, ranges, distributions  Apply your sniff test Tip #6: Unleash your curiosity
  14. 14. Use engineering standards Make your work share-able & re-usable:  Source Control & Notebooks  Data dictionary  Data contracts & SLAs  Privacy, Compliance, Ethics Gather feedback to improve your results:  Peer & Code reviews  Office hours & brownbags  Stakeholder presentation & action  Publish  Retrospectives Tip #7: Role model quality approaches
  15. 15. Data Science in the real world  Causation vs correlation  Experiment considerations (time to market, opportunity cost, ethics)  Done (and simple) are better than perfect  80/20 rule  Model explain-ability  Skewed populations  Value of data quality, monitoring and improving data sets Tip #8: Prioritize practicality
  16. 16. Data Science Lifecycle Problem & Hypothesis Problem Framing Data Acquisition & Exploration Analysis & Predictive Modeling Evaluation & Reviews Deployment & Socialization Data Science Lifecycle
  17. 17. Scientists must speak  Presentation skills  Be concise  Focus on the takeaways  Connect with your audience  Use volume, eye contact, pauses  Practice Tip #9: Land your message
  18. 18. Simplify for Impact  Quiz: Which one is better? Tip #11: Eliminate Distractions
  19. 19. Leverage Libraries Credit: Cole Nussbaumer Knaflic
  20. 20. 21 Grow your career Hone your approach • Become a SME • Deliver results Increase your impact • Transform a space • Share new ideas • Help & represent the team Expand your horizons • Mentoring, Network • Books, Courses, Events • Company & Industry
  21. 21. Stay in Touch https://www.linkedin.com/in/cohenlisa/ https://medium.com/data-science-at-microsoft Lisa.Cohen@microsoft.com
  22. 22. 23 Q&A
  23. 23. © Copyright Microsoft Corporation. All rights reserved.

Editor's Notes

  • Careers are only built in retrospect
    You can’t connect the dots looking forward; you can only connect them looking backwards. So you have to trust that the dots will somehow connect in your future. You have to trust in something — your gut, destiny, life, karma, whatever. – Steve Jobs
  • Org maturity levels
  • Tips:
    Don’t feel limited by the boundaries
    Follow your passion and leverage your strengths
    Share your interests with your manager
    Take on projects that align with your future goals
  • Tips:
    Make a plan for career experiences & learnings
    Tackle Imposter Syndrome
    Apply the Venn diagram to the organization or Data Science field

    Notes:
    Leverage the diversity of the team backgrounds, with group projects
    Have fun with the team. Help & contribute to each other.
  • Pros/Cons & Tips:
    Pro: Drive product direction, product management
    Con/Tips: Join DS communities, Find a mentor
    Pro: More career paths, diverse projects, like-minded peers & team projects
    Con/Tips: Steering meetings, Joint planning, Join aliases for context, Find champs

    Notes:
    Grass is greener.
    All exist at Microsoft. MS is centralized at the product level
  • Data Science maturity stages
    Partner vs serving ad hocs, bring new ideas, file work
    Saying no, moving replies out of email, office hours
  • https://www.google.com/search?q=arc+arrow&tbm=isch&hl=en&chips=q:arc+arrow,g_1:blue:AlXoYtHtNkI%3D&rlz=1C1GCEU_enUS820US820&hl=en&ved=2ahUKEwjalJPLzfTpAhUzkZ4KHbMSAZIQ4lYoCHoECAEQJQ&biw=1479&bih=2261#imgrc=RX46V2bKlXmNmM&imgdii=wHBImvC5s7jK6M
  • https://unsplash.com/photos/M-EwSRl8BK8
    Bring together end-to-end datasets
    Gain context from source owners
  • Use visual aids, make your message pop
    Make your visuals work for you, not against you
  • https://pixabay.com/photos/raise-challenge-landscape-mountain-3338589/
  • https://www.flaticon.com/free-icon/linkedin_174857
    https://medium.com/@Medium
  • ×