Daliana Liu

Today, an intern at HBO made a huge mistake.

Is the intern to blame? Making mistakes is part of being human, and what's important is to create a robust system to catch those errors.

What mistake have you made as an intern?

p.s. I write more about career and data science in my exclusive newsletter, subscribe here: www.dalianaliu.com

Can AI can predict soccer goals?

Watch the video to find out. If you are interested in the model we used to build the 'goal predictor', check out this blog we collaborated with Sportradar:https://lnkd.in/gAjC6hU

What's your favorite AI use case for sports?

#datascience #ai #machinelearning

A complete table of Generative AI products (with 3 categories):
1. Extraction:
image/video => text
Analyze existing images to generate insights

2. Generation:
text => text/image/video/code
Create new content based on simple text input

3. Enhancement: image, video, text, code
Making existing content better, autocomplete

*I share my about data science and career at www.dalianaliu.com

#datascience #machinelearning

Image credit: https://lnkd.in/g_3H_QaJ

I’m glad this is happening! After it’s approved, you can quote this law next time when an employer doesn’t disclose the pay band. For a specific level, you can also use levels.fyi, teamblind to find out.

*I share more about a career in data science here:www.dalianaliu.com free to subscribe.

Want to learn more about how Machine Learning is applied in companies like Amazon, Lyft, Airbnb and DoorDash?

This is a list of 100+ papers and blogs on how companies are solving forecasting problems, recommendation systems, search and ranking, computer vision, etc.

https://lnkd.in/gAJ36y9

(Curated by Eugene Yan applied scientist at Amazon.)

Listen to my podcast “The Data Scientist Show“ for more tips on machine learning and career: https://lnkd.in/gdqgbxPZ

#datascience #machinelearning

Want $1,000,000s worth of machine learning knowledge for free?

Companies pay six figures to host challenges on Kaggle, after the competition closes, many Kagglers like Bojan Tunguz, Ph.D. & Mark Tenenholtz will post their solutions - free and well documented.

What's better, Farid Rashidi created an easy web app finding all the top solutions for challenges with search functions: https://lnkd.in/eK5u8VFt

*I saw this from Jesper Dramsch, PhD's post. Learn more Kaggle tips and tricks from Jesper:
Apple: https://lnkd.in/e-UqKvXN
YouTube: https://lnkd.in/e722PzCc
Spotify: https://lnkd.in/evJt4W8V

#machinelearning #datascience

Don't let the lack of math skills stops you from getting into data science.

The blog below listed some essential skills. I like that it points out a lot of calculations can be done using Python/R libraries.

Here is how I would prioritize:
- I would spend more time on statistics, linear regression, and probabilities.
Because they are most commonly used concepts and the foundation to solve most problems.

- Learn the needed calculus and linear algebra when it's needed.
Unless you are doing machine learning research, you only need to know the basics to learn machine learning. They are important, but learn them as you get into machine learning, and don't need to get too deep into it when you get started.

More explained in this blog below by Benjamin Obi Tayo
https://lnkd.in/gyykz5Gs

#datascience #machinelearning

You don't need a background in AI to be in AI.

A good foundation in engineering goes a long way.

Do NOT be a data scientist if you:

1. are easily frustrated when research doesn't yield results.
2. don't like to deal with vaguely defined problems.
3. only love the math and theory, but don't want to communicate with non-tech folks.

Note that I didn't say 'bad at math'. Everyone can learn tools to do math.

But to be a data scientist, you need the mindset to deal with ambiguity and uncertainty, so you can solve the business problem.

What are some other qualities you believe a data scientist must have?

Follow my newsletter www.dalianaliu.com for more exclusive take on data science and career.

#datascience #career

I used to feel guilty about not reading or learning something new over the weekend. Even when I was at a party, I couldn't fully enjoy myself and felt present.

I thought it's because the education system made me feel the stress of doing homework

But then I see my peers who grew up in similar environment but still able to have more fun.

So I realized it's myself didn't give me the permission to live, what's the point of chasing and being productive everyday, for what?

I don't want to die and realized that I never lived.

The best engineer/scientist I know are those who do great work and also know how to enjoy themselves.

You are not wasting your time if you want to slow down and take a break.

“You waste years by not being able to waste hours”

Give yourself a few hours, a few days to 'waste'. It's not wasting, it's called 'living'.

Is autoML eating data scientists' jobs?

Yes and no.

Yes - if all you do is the 'sexy' part of data science: training different models, tuning parameters, picking the best-model based on performance metrics, you should be worries because those tasks are getting automated.

However, if you have been doing a lot unsexy 'dirty work' like cleaning data, selecting data, model testing, scaling models in production, autoML can make your work more efficient. And your experience in the end-to-end data science workflow makes you more valuable.

The unsexy data science work is hardest to get automated.

Why? Because it involved human decisions.

For example:
• whether you should build a pipeline to collect new some data that has 50% chance of being useful? (data selection decisions)
• is the model good enough to launch now with 85% accuracy, or do we need 90% and wait for 6 months? ( trade-offs)
• is the model working well with other features in the product, or it's hurting the customer experience? (product sense)

Honestly, most data scientists are already doing the dirty work, but not a lot of us enjoy doing it and want to avoid it.

It's time to change the mindset - we are not just modelers, we are problem solvers. The data science work begins with understanding the problem and ends with generating impact.

A few things to start with:
• Challenge assumptions
• Learn domain knowledge
• Know when to make trade-offs
• Talk to product managers and engineers
• Get comfortable with non-DS related challenges

The future of data science is to use automation tools well and get good at the things that can't be automated.

*This is inspired by my conversation with Greg Tanaka, learn more about the future of data science, forecasting, and quantitative trading on 'the data scientist show':
Apple:https://lnkd.in/ga-KVEvY
Spotify:https://lnkd.in/gZ5ESM3J

#datascience #machinelearning

I don't know why but this month feels like the 24th month of 2020.

And... I am tired.

• I'm tired of the zoom meetings & virtual “happy hours“.
• I'm tired of sensational news titles, “is AI gonna kill us?“, “Is data science dying?“
• I'm tired of the Linkedin polls and the never-ending debate of whether one tool is better than the other (oh, Python or R?)
• I'm tired of working in my living room and living in my office.
• I'm tired of waking up to the same room, same routine.

I want to take a break.
A long break.

For previous holidays, I used to work on a passion project, set goals, and read books. But I just want to be lazy this year.

If you feel the same, give yourself permission to do NOTHING.

You worked hard.
You deserve a break. A real break.
You'll feel more productive afterwards.

You can start your 2022 on your own term, even if it's a few weeks later.

Hope everyone get rested❤️.

-- Daliana Liu

A free course for you to get into data science:

Data Science 365 just offered unlimited access to their platform for two whole weeks 100% free of charge. No credit card is required for registration. (I hate those “free“ registrations that require credit cards...)

I like this offer because they also provide real data science projects along with the courses, so you know how to apply the skills in the real world. It's from Nov 6 to Nov 20, and you can register here: https://lnkd.in/eiRTeP2C

*I share more about data science career at www.dalianaliu.com free to subscribe

#datascience