rockyourdata

Please post your reply here

Aptitude: curiosity @IBMbigdata

Skills: analytical - gotta love the numbers @IBMbigdata

Clearly, a data scientist must be good with numbers: stat, math, probabilities.But they should also be adept at critical thinking: distilling problems down to variables that might be measured and tested through statistical means.

curiosity and patience, but also an open-mindedness when it comes to methods and programming languages.

A data scientist needs to be a good programmer

this is a great question one that is always evolving and me thinks will continue to evolve.

A data scientist should be @IBMbigdata critical, curious and patient.

The ability to answer creative,crucial,interesting,insightful questions, combined with sufficient technical and programming skills.

being able to tell when the results don't makes sense is pretty important when it comes to data science. You're the (human) adult in the room.

Important is to know about the business @IBMbigdata

A healthy dose of skepticism, readiness to question the status quo, and sensitivity to the reality that it's not just "data", it's people whose lives will be affected by data scientists' decisions

Here are the areas with which data scientists must become familiar: linear algebra, basic statistics, linear and logistic regression, data mining, predictive modeling, cluster analysis, association rules, market-basket analysis, decision trees,

Nible to get the business context and Solution

A data scientist has to be skilled at "feature engineering"--ie., identify indep & dependent variables) to build into stat models. And they should have an agile mind for adjusting the feature mix as they build/test alternative models

Data wrangling either with writing some basic code or an interface - managing data is key

. @IBMbigdata i've predicted that data science will move to analysts and programming won't be required in the future (soon)

A good data scientist should never love stat, math and data for its own sake, but rather as a means toward the end of ascertaining correlations and other patterns that it all may reveal.

A data scientist has to have an enquiring and creative mind and be highly driven

Data Scientist= Data structures manipulator + Scientific degree holder

Time for question #3 - take a look at the top of your screen

@jameskobielus cofounder Jeff Hammerbacher of Cloudera had the best line comparing athlete to data scientist; love for the sport is what makes someone great

#techathlete in data science is like "gym rat" or "data rat" always hanging around never leaving the gym :-)

I don't see question #3 at the top of my screen.

We have seen a rapid move to self-serve analytics @furrier However, using those tools without the skills/knowledge is not very useful. Exhibit #1: Excel

Please post your replies here

The bench used by Tony Stark!!

I love this topic

The ideal #datascience modeling workbench should incorporate Spark for modeling, a deep statistical algorithm library (machine learning, NLP, etc.), support for R, Scala, and Python development, and access to a hybrid "data lake"

. @IBMbigdata workbenching notion is changing fast esp now that the required updates to data sets can be real time; API aggregation data wrangling.. a new CMS or tooling is needed big time- I live this problem every day

My wishlist: team communication tools, shared and private repositories of data, source control for data and models, and while we are at it, with lots of memory and storage capabilities @IBMbigdata

The ideal #datascience modeling workbench would automate most of the data discovery, acquisition, and preparation tasks, enabling modelers to focus on fast model development, testing, and deployment.

I see data integration plus WSWYG tool being really big oppty

Other than Jarvis: Prototyping language(Python or R), ML library with distributed computing(@Spark), Visualizations(@Bokeh), IBM Watson!

It should be dynamic in nature and visual as much as it can. This ensures greater understanding of the approach.

R, Python, Spark are essential for sure, but I also find a good notebook and a pen are often forgotten tools by even the best modelers.

.@hungry_thalamus ...and colored markers too!

Time for Question #6

Workbench solutions for big data I'm very interested in; if you get more info please send it to me

I'd love to do a workbench crowdchat topic - go deep in the weeds on how to

R, Scala, python, machine learning library and other deep statistical library such as NLP

Please post your reply here

waiting for your valuable inputs

Hackathons!!

@jameskobielus Here you go

Greta answer @anshmaniaa

A number of modelling approaches to handle high leverage observations or violation of assumptions. Models can be both parametric (Ex regression with Huber, White, Sandwich variance estimators) as well as non-parametric(e.g. quantile regression)

Start from your current power analysts in various business units: marketing, ops, etc. There's probably a lot of statistical analysis talent, coupled with deep domain expertise, that you can nurture into high-powered data scientists

Look inside your company before recruiting from outside, The people within have already met certain criteria

in 2016, it's easy to spot good modelers with what they've already done. Cultivating them requires teamwork with developers, engineers, and people in the field gathering data. Also, they need the freedom to explore new methods.

Explore various data-scientist engagement tactics--centers of excellence, online challenges, contests, hackathons, community events--to see what hot talent shows up and what type of work they can do.

I highly recommend access to participate in Data4Good projects

Look for university graduates, they are eager to get started and have received the latest advice and training

Establish internship programs in which smart people from a wide range of higher-educations institutions and programs can show what they're capable of. They'll bring passion, commitment, and fresh ideas to your data science practice

Great points @jameskobielus @lisabriercliffe @DrZeydy

Don't fall into the trap for thinking the best and brightest data scientists are those who come from the "top universities." Statistical smarts are probably distributed across the entire higher-education landscape.

Short term: Look for candidates in non-traditional places, and follow the model of GHCQ’s recruitment puzzles.
Longer term: Let kids be kids. Allow them to fail so they develop resilience and self esteem. Nurture their curiosity and imagination so they grow to become creative, confident and engaged thinkers.

In a corporate environment, I recommend job rotation

Time for question #4 - please look at the top of your screen for Question #4

Curious minds with aptitude for data and analytics especially young college hires given jobs shadowing