The Data Prophet

Simplifying Data. Creator of predictive models.

[The Data Prophet @mldataguy]

ULITMATE DATA MASTERY PLAYBOOK

This is the breakdown of everything data that will make you successful now & later.

Highest level metadata of #data posts.

A🧵of 🧵's

[The Data Prophet @mldataguy]

5) They have been targeting Ukrainian water, gas and electricity connections in an attempt to freeze and starve the civilian population into surrender.

[The Data Prophet @mldataguy]

The Battle of Bakhmut is shaping up to be one of the bloodiest and dirtiest of the entire war.

[The Data Prophet @mldataguy]

4) My bet is that the Russians will partially surround Bakhmut during the winter and force a Ukrainian retreat.

[The Data Prophet @mldataguy]

Much of data revolves around visualization to an end user,

But data scientists/engineers need visualizing tools as well.

While users want visuals that give answers, data scientists need visuals that create questions.

Data scientist need visuals that depict the problem

[The Data Prophet @mldataguy]

What they don’t tell you about data science is that every project you do is a sales pitch.

You have to sell to get the project off the ground.

Sell the progress being made.

Then you have to make the final sales pitch to prove why your model should go to prod.

[The Data Prophet @mldataguy]

Want to optimize #data queries & processes.

Less is always more.

Less tables, less columns, less joins, less data in memory.

Always try to narrow down first

[The Data Prophet @mldataguy]

My brain has a very hard time specializing in one area. The connections to everything else is what drives insight for myself. Data has created a lens with which everything flows

[Patrick Wendell @pwendell]

Unfortunately for all of us, this is a highly accurate depiction of #MLOps. @Databricks is focused on making this much simpler.

From: "Machine Learning Operations (MLOps): Overview, Definition, and Architecture". Dominik Kreuzberger, et al https://t.co/QGj1djER5v

[The Data Prophet @mldataguy]

@ZainullAbiddinD Lol exactly

[The Data Prophet @mldataguy]

A quick dirty way is to setup a smtp email function in your prod environment that you can call to send emails with parts of your code fail.

[The Data Prophet @mldataguy]

When you build a house, you make sure to install your plumbing first. It wouldn’t make sense to install a shower after playing in the dirt.

Yet this is exactly what we do in data ecosystems.

We build monitoring systems after data systems begin to fail.

[The Data Prophet @mldataguy]

So what should you do?

Adopt a mindset that pipelines can only go into production if they have monitoring & unit testing put into place.

Services like datadog allow a quick way to do so & set up email notifications.

[The Data Prophet @mldataguy]

Feeling low on energy this morning?

Drop a production table & feel the rush of adrenaline pump through those veins.

#data

[The Data Prophet @mldataguy]

The real value of well modeled data is better decision making.

[The Data Prophet @mldataguy]

Good Data > Big Data

Every. Damn. Day

[The Data Prophet @mldataguy]

@_Lotus77 @WifeyAlpha Down for it

[The Data Prophet @mldataguy]

Why fraud always grows & is exposed at end of market cycles.

Excerpt From
Lying for Money
Dan Davies

[The Data Prophet @mldataguy]

“since pricing in a competitive industry tends to be set by the lowest bidder not the smartest bidder, there will always be pressure to get a short-term cost advantage by economising on the kinds of bureaucratic, not directly productive checks and balances that pick up fraud”