Downloaded 10 times
![WHAT IS A DATA
SCIENTIST’S ROLE IN
AN AGILETEAM?
“The role of the artist [data scientist] is to ask questions, not [and] answer them.”
― Anton Pavlovich Chekhov
15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 15](https://image.slidesharecdn.com/doing-data-science-the-agile-way-181015153715/75/Doing-Data-Science-the-Agile-Way-15-2048.jpg)
![References
■ [Wikipedia: Data Science]
■ [Forbes: Data Science:What'sThe Half-LifeOf A Buzzword?]
■ [Panaseer.com:Can Data Scientists Sprint?]
■ [Episode 21: Learning by Failing:TroubleshootingAgile]
15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 34](https://image.slidesharecdn.com/doing-data-science-the-agile-way-181015153715/75/Doing-Data-Science-the-Agile-Way-34-2048.jpg)
More Related Content
Similar to Doing Data Science the Agile Way
Doing Data Science the Agile Way
- 1. DOING DATA SCIENCE THE AGILEWAY by Dr. Christos Hadjinikolis Senior Consultant | ML Engineer | Data Reply UK
- 2. HOW DOES ADATA SCIENTIST FIT IN AN AGILETEAM? &WHAT IS HIS ROLE?
- 3. A few words aboutme … ■ Completed my Doctoral Studies in 2015 ■ Expertise in Graph Analytics ■ Started working as a Data Science Consultant at Data Reply UK soon after ■ Short term roles (2-3 months) – Data Engineering – Data Science – Industries: ■ Finance ■ Retail ■ Government ■ … but, for the last 1 ½ year I have been working as part of a Digital Transformations team …
- 4. … it wasdifferent Usual scenario: ■ Company has a problem ■ Consultant goes in ■ Consultant comes out (problem solved) ■ …on to the next challenge Not this time … I found myself becoming part of an Agile Software DevelopmentTeam and … it was confusing.
- 5. Overview ■ What isa Data Scientist? ■ What is Agile Software Development? ■ What is a Data Scientist’s Role in an agile team? ■ Creating Experiments NotTasks ■ Setting Clear Expectations to Stakeholders ■ How shouldWork be Split? ■ Final Remarks 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 5
- 6. WHAT IS A DATASCIENTIST? “... Accept no one’s definition of your life; define yourself.” ― Robert Frost
- 7. What is aData Scientist? 1960 Peter Naur, using it as: “…a substitute for Computer Science”! 1996 at the IFCS meeting in Kobe on: ”Data Science, Classification, and Related Methods” 1997 C.F. JeffWu: “Statistics=Data Science?” 2007 Jim Gray: “Data-driven science is the 4th paradigm of Science” 2012 Harvard Business Review: “Data Scientist:The Sexiest Job of the 21st Century” 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 7
- 8. What is aData Scientist? ■ Illinois Institute ofTechnology, Shlomo E. Argamon – These people need to have strong : ■ statistical knowledge, ■ software engineering ability, and; ■ communication skills. ■ Forbes, Gil Press – “… is a buzzword without a clear definition and has simply replaced business analytics” 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 8
- 9. What is aData Scientist? “A professional that uses: ■ scientific methods ; ■ processes; ■ algorithms and; ■ systems, to extract insights from structured or unstructured data.” * Data Science,Wikipedia: https://en.wikipedia.org/wiki/Data_science 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 9
- 10. What is aData Scientist? Data Scientist : DATA INFORMATION ~KNOWLEDGE Data Scientist + Domain Expert … INFORMATION KNOWLEDGE WISDOM 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 10
- 11. WHAT IS AGILE SOFTWARE DEVELOPMENT? Inshort, it’s an iterative and incremental development method …
- 12. What is AgileSoftware Development? ■ An Approach to Software Delivery – Time Boxed – Iterative – Incremental ■ Projects are Split into Stories – User Functionality Components – They are Prioritised – Continuously Delivering them in Short Cycles (~1/2 weeks) 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 12
- 13. The Manifesto ofAgile Software Development Values (left preferred over right) ■ Individuals and Interactions over processes and tools ■ Working Software over comprehensive documentation ■ Customer Collaboration over contract negotiation ■ Responding to Change over following a plan 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 13
- 14. Why Agile SoftwareDevelopment? Given a Constantly Evolving Functional and Technical Landscape: ■ Maintain a Focus on the Rapid Delivery of Business Value ■ Adapt to Changing Requirements ■ Reduces Risk of Overall Failure – By adopting a “fail fast, fail often” mindset – Rapid end-user feedbacks help better direct focus 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 14
- 15. WHAT IS ADATA SCIENTIST’S ROLE IN AN AGILETEAM? “The role of the artist [data scientist] is to ask questions, not [and] answer them.” ― Anton Pavlovich Chekhov 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 15
- 16.
- 17. … but one day… Let’stake this into production • Create a solution for automated insights generation • Create some tasks in JIRA • Get to work …
- 18. … confusion The overallapproach was not going very well • Ill-defined Sprints • Far from an early deliveries • More Scope • A never ending Story … On top of these … • Existential Crisis • Who am I? ?
- 19. What is aData Scientist’s Role in an Agile Team? “If practicing Agile Software Development produces value from learning through failing fast and safe … then a Data Scientist’s role in an Agile team is to help the team to fail faster!” 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 19
- 20. What is aData Scientist’s Role in an AgileTeam? ■ Ask Questions and then Answer those Questions! ■ Iterative HypothesisTesting ■ It’s about “Experiments”! 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 20
- 21. CREATING EXPERIMENTS & NOT TASKS “Don'tcry because it's over, smile because it happened.” ― Dr. Seuss
- 22. … back tothe story Reasons behind our lack of efficiency • Too many assumptions in place • Incomplete understanding of the data • Moved from a successful data analysis to product development • “Many more hypotheses had to be validated in order to bring a sound and complete analytics platform to life”. !
- 23. Creating Experiments ¬Tasks ■ Experiments … that is what you add in your backlog ■ A successful scientific experiment requires clear and accurate hypothesis – Just the thought process of defining it helps! ■ Bad hypothesis will always lead to a real failure. ■ The goal is not to confirm an assumption – … but learn through the process of validating it! ■ Research is key! 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 23
- 24. SETTING CLEAR EXPECTATIONS “Insanity isdoing the same thing, over and over again, but expecting different results.” ― Narcotics Anonymous
- 25. Setting Clear Expectations ■The term “experiment” carries negative connotations ■ Managers are trained to avoid them, when they should be encouraging it ■ Experimentation, tolerates failure as education – It’s not about failing – It’s about learning ■ Experiencing backpressure is inevitable … 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 25
- 26. Cultivating the Right Culture ■Courage to be uncertain ■ Stakeholders need to understand the value of experimentation ■ Loss aversion will eventually kick in; usually very early! ■ … but, this is 100% natural human behaviour
- 27. SPLITTINGWORK Someone needs toadd a smart quote here! Assignee: Christos
- 28. SplittingWork ■ Why areyou researching this? – Identify your objectives – Prioritise your objectives ■ For each objective: – What questions do I need to ask and answer? – Create a list of experiments – Group relevant work together in the form of sub-tasks There you have it; a data story is born! 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 28 * Panaseer.com: Can Data Scientists Sprint?
- 29.
- 30.
- 31. Final Remarks ■ AgileSoftware Development is not a panacea ■ Data Scientists + Agile – When a team struggles with data … then yes! – When assumptions need to be validated! ■ An experiment-friendly culture needs to be cultivated! 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 31
- 32. Final Remarks “It’s notabout doing it in the best possible way, but, simply and more importantly, about doing it!” Jeffrey Fredrick & Douglas Squirrel 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 32
- 33.
- 34. References ■ [Wikipedia: DataScience] ■ [Forbes: Data Science:What'sThe Half-LifeOf A Buzzword?] ■ [Panaseer.com:Can Data Scientists Sprint?] ■ [Episode 21: Learning by Failing:TroubleshootingAgile] 15/10/2018 Dr Christos Hadjinikolis | Senior Consultant - ML Engineer | Data Reply Uk 34
Editor's Notes
- #2 Hello everyone. I am Christos; very happy to be here today. Thanks for attending my talk ... So, the topic of my presentation is “Doing data science the agile way” and I want to start with explaining what it is exactly that I will be discussing today.
- #3 So… here is what I want to get across: “How does a Data Scientist fit in an Agile team and what is his role” This is both a slide intended for setting the scene for what is to follow but also for clarification purposes—admittedly the title might be a bit misleading in the sense that this talk could be just as well about a team of data scientists working together which is not. In fact ... it is a bit of a story about how I have been working over the last 3 years or so ...
- #4 …. A few words about me :) I know I don’t come across as a hipster but I felt like it would be a lot more cheesy if I had put a picture of myself in there! So …
- #6 … so to tell you this story there is a number of things I want to cover … Here is our agenda then for the day. We will start with discussing Data Science and Data Scientists in general and move to discussing their role, as I see it, within an agile team. We will then move on to discussing: Setting Clear Expectations to Stakeholders Adding Experiments in the Backlog (Not Tasks) Leveraging Value from Intermediate Results How should Work be Split? While I go through these different topics I will be narrating personal experiences that got me to where I am today, which is standing here on this stage and talking about Data Science with you. Note: Assigning means Project Manager assignes But in agile we are talking about self organised teams ... anyone chooses from the backlog Pulling from the backlog
- #8 The term has appeared in various contexts over the past thirty years but did not become an established term until recently. As early as 1960, Peter Naur was a Danish computer science pioneer and Turing award winner, using it to refer to computer science. He later introduced the term "datalogy” and eventually data science was used to refer to contemporary data processing methods that are used in a wide range of applications In 1996, members of the International Federation of Classification Societies (IFCS) met in Kobe for their biennial conference. Here, for the first time, the term data science is included in the title of the conference ("Data Science, classification, and related methods") Founded in 1985, the IFCS is a federation of national, regional, and linguistically-based classification societies. It is a non-profit, non-political scientific organization, whose aims are to further classification research. In November 1997, C.F. Jeff Wu gave the inaugural lecture entitled "Statistics = Data Science?” Chien-Fu Jeff Wu is the Coca-Cola Chair in Engineering Statistics and Professor in the H. Milton Stewart School of Industrial and Systems Engineering at the Georgia Institute of Technology. Around 2007, Turing award winner Jim Gray envisioned "data-driven science" as a "fourth paradigm" of science that uses the computational analysis of large data as primary scientific method In the 2012 Harvard Business Review article "Data Scientist: The Sexiest Job of the 21st Century",DJ Patil claims to have coined this term in 2008 with Jeff Hammerbacher to define their jobs at LinkedIn and Facebook Dhanurjay "DJ" Patil is an American mathematician and computer scientist who served as the Chief Data Scientist of the United States Office of Science and Technology Policy. from 2015 to 2017. Jeff Hammerbacher is a data scientist as well as chief scientist and cofounder at Cloudera. Before you know it international conferences start appearing, graduate degrees and MScs…
- #9 Some people, like Shlomo Engelson, that: Skills that were assumed to be possessed by a Data Scientist can only be achieved through a well-designed multidisciplinary program. This distinct body of knowledge and skills cannot be conveyed by traditional educational programs such as statistics or computer science. Shlomo Engelson Argamon is the director of Illinois Institute of Technology’s new Master in Data Science program Gill Press Managing Partner at gPress, a marketing, publishing, research and education consultancy
- #10 I will go with something a lot more generic and simple today … notice … I didn’t say anything about interpreting those insights
- #11 Someone that will get you from data to information … notice … I didn’t say anything about interpreting those insights
- #13 Agile is a time boxed, iterative approach to software delivery that builds software incrementally from the start of the project, instead of trying to deliver it all at once near the end. It works by breaking projects down into little bits of user functionality called user stories, prioritizing them, and then continuously delivering them in short two week cycles called iterations.
- #14 All these are better described in the Agile Manifesto. A document put together in 2001, seventeen software developers met at a resort in Snowbird, Utah which also documents a set of values and principles That is to say, the items on the left are valued more than the items on the right. As Scott Ambler elucidated:[19] Tools and processes are important, but it is more important to have competent people working together effectively. Good documentation is useful in helping people to understand how the software is built and how to use it, but the main point of development is to create software, not documentation. A contract is important but is no substitute for working closely with customers to discover what they need. A project plan is important, but it must not be too rigid to accommodate changes in technology or the environment, stakeholders' priorities, and people's understanding of the problem and its solution. Agile software development principles[edit] The Manifesto for Agile Software Development is based on twelve principles:[21] Customer satisfaction by early and continuous delivery of valuable software Welcome changing requirements, even in late development Working software is delivered frequently (weeks rather than months) Close, daily cooperation between business people and developers Projects are built around motivated individuals, who should be trusted Face-to-face conversation is the best form of communication (co-location) Working software is the primary measure of progress Sustainable development, able to maintain a constant pace Continuous attention to technical excellence and good design Simplicity—the art of maximizing the amount of work not done—is essential Best architectures, requirements, and designs emerge from self-organizing teams Regularly, the team reflects on how to become more effective, and adjusts accordingly
- #15 The acceptance of “fail fast, fail often” mindset ensures the iterative approach to software delivery, where rapid customer feedback loops help the developers better direct their efforts and deliver the features the customers really expect. Go for the MVP (minimum viable product)
- #17 So there I was 1 and half back when I first joined my current team … Analyzing my data working my way through my python notebooks Generating my visualizations Discussing insights and outputs with domain experts in the company Life was great !
- #18 We have some good results … lets take this to production Put together an architecture with the team of what is needed Do a bit of planning, create some stories … put some tickets and get cracking
- #19 Now I believe that I am a competent Engineer; I had the support of the team as well so the software engineering bit was not much of a problem. The think is that I found myself in a position where I couldn’t really identify what was my role in the team … The overall approach was not going very well Sprints were not producing the desired output We were far from an early delivery with every new iteration More scope was being introduced every time It seemed like a never ending story … On top of these … Existential Crisis Who am I?
- #20 https://twitter.com/johncutlefish/status/1044410010237726721 Agile Principles Are people familiar with everything? Fail fast; fail safe ... Learning fast! Fast feedback Working together with Software Engineers If you have a team that is always struggling with data then a data scientists fits in !
- #21 Does anyone know how PhDs work? In fact how many of you have a PhD ? Great you all have my sympathy! I know what you 've been through! Here is how it works ... Ask Questions and then answer those Questions! Iterative Hypothesis Testing “Experiment” also requires a hypothesis - a lot of projects don’t have a clear and understood hypothesis to base experiments off. Generally Fail Fast Data Scientists “Lets fail faster!”
- #23 Relied on a number of assumptions that were not valid No clear understanding of the
- #24 Experiments … that is what you add in your backlog … not tasks A successful scientific experiment requires clear and accurate hypothesis I have observed teams get extremely bogged down on formal hypothesis statements (and null hypothesis, etc.) They miss the spirit of inquiry IMHO. It is the thought process that counts. Many teams don’t accurately define (or even) identify the problem they are trying to solve with their experiments. The goal is not to confirm an assumption … but through the process of validating it! Research is key!
- #25 PMs spending countless hours “influencing stakeholders” ... but not having time to talk to customers/users and be available for their team.
- #26 The term “experiment” carries negative connotations Managers are trained to avoid them, when they should be encouraging it Experimentation as a path to discovery tolerates failure as education It’s not about failing It’s about learning
- #27 It’s about cultivating the right Culture Not enough courage to be uncertain, to throw some work out ... Also, teams struggle if higher stakeholders do not get the value of experimentation, rarely the other way around. Loss aversion kicks in very early, and expands to encompass every single person involved with testing a new idea including dev, product, stakeholders, etc,. It requires a huge cultural effort to combat this due to perfectly natural human behaviour.
- #28 https://panaseer.com/2016/03/21/can-data-scientists-sprint/
- #29 Ask yourself, “Why am I doing this research?” If there are multiple research goals, find out which are the most urgent (according to the current prioritisation in your business) and limit the research scope accordingly. If you have multiple research tasks you many need to pick a few goals from each, or focus on one, depending on the prioritisation. For each high-priority goal, consider what questions you need to answer in order to achieve it. Write a to do list, based on the goal and associated questions, as if you were just planning your own work. Review the list and try to group related points together. Think of a higher-level summary for each of these groups – this is your Jira ticket. The finer details are your sub-tasks. Rather than completing half of it in detail, consider making a first pass using simple solutions and then using another sprint to revisit the issue and build in complexity. Having a sprint demo scheduled at the end of the week encourages this way of working as you’ll need something coherent and self-contained to present.
- #33 by Jeffrey Fredrick and Douglas Squirrel