Naked Data Science

A Technology podcast
Good podcast? Give it some love!

Best Episodes of Naked Data Science

Mark All
Search Episodes...
When most people think about data science, they have some sort of Machine Learning in mind. But the truth is many data-intensive problems don't need Machine Learning, even in big tech companies like FAANG. In this episode, Nima will share the reasons why he went from a researcher in Machine Learning to become a data-driven problem solver and give a couple of tips on how you can make that transformation too. BTW, we are trying out a new format for our podcast based on audience feedback. This episode is shorter than previous episodes and focuses on a more specific topic. Enjoy.If you like this episode, you will like our insider's guides on how to solve data science problems like a detective. We are also sharing new materials and training every week. Get these free insider's guides today. 
If you are still scrolling through your Jupyter notebook when presenting your data science work, you are not giving your work the attention it deserves. And when I say it probably even limits your salary and career, it is not exaggerating. In this episode, we will show you why presenting is not window-dressing, but a key problem-solving skill in data science. We will give you seven practical tips and a presentation template that can drastically improve your next presentation.  If you like this episode, you will like our insider's guides on how to solve data science problems like a detective. We are also sharing new materials and training every week. Get these free insider's guides today. 
There were cognitive biases in the data science work you did. And there will be more cognitive biases in all the future work you will ever do. They are just part of being human. But if you don't pay attention to HOW these cognitive biases affect your work, you can easily waste weeks if not months chasing after the wrong things. In this episode, we will talk about some common cognitive biases that affect the data science work, and how you can deal with them. By the way, if you like this podcast, you will like our free guides. We take the most popular topics from past episodes - this can be fixing projects that are not going well, receiving the recognition you deserve, building intuitions on different types of models and machine learning methods. We condense each topic into a short PDF, which you can use as a quick reference in your daily work to practice skills and develop strong intuitions on these topics. You can get these free guides at www.nds.show.
What happens when you are not working on interesting work? It is boring, you feel stuck, and your skills and career stop developing. But it is also very bad for your company: they now have an employee who is not delivering good outcome while still requiring high effort to manage, So obviously, it would be great if you and your company can always find work that is interesting to you. In this episode, we are going to show you some simple ways to do exactly that.By the way, if you like this podcast, you will like our free guides. We take the most popular topics from past episodes - this can be fixing projects that are not going well, receiving the recognition you deserve, building intuitions on different types of models and machine learning methods. We condense each topic into a short PDF, which you can use as a quick reference in your daily work to practice skills and develop strong intuitions on these topics. You can get these free guides at www.nds.show.
Unless you have been living in a cave in the past 2 years, you have heard of AutoML. And depending on where you have heard it from, it can be the best thing ever happened to data science, the evil invention that will put thousands of data scientists out of their jobs, or anything in between. In this episode, we talk about the state of the art AutoML, what is hype versus what is reality, how to think about it practically, and how you can get started with AutoML in your team. If you like this podcast, you will like our free guides. We take the most popular topics from past episodes - this can be job hunting misconceptions for data scientists, top mistakes in data science team communications, understanding AutoML in plain English, and many other topics. We condense each topic into a short PDF, which you can use as a quick reference in your daily work to practice skills and develop strong intuitions on these topics. You can get these free guides at www.nds.show.
What can you do about about the ethics of AI, Machine Learning, and other data science solutions in your daily work. Why it is important to think about implications first, not technologies. The four principles we use to address ethical challenges. Some practical ethic codes for data scientists.
Why data science team communication is so difficult. Analytics Translator is not the solution. Role of PM in a data intensive solution team. Why you shouldn't rely on everyone's notes. What to do when you receive a long text. When to put things in writing and when not to. Handling difficult conversations.
Systems thinking to make sense of your data science work. Similarity between dead fishes and recommender systems. Effect of time and feedback loop on your models. Look beyond your dataset. Applying systems thinking to people and teams. How to change a system without breaking your back.
How domain knowledge can supercharge your data science work. The half life of truth at three levels of business domain knowledge. Why it is important to follow the money in data science work. Three ways to acquire new domain knowledge fast.
How thinking in questions can help you communicate your work effectively, especially to non-data-scientists. Avoid getting lost when finding your path to a solution. Three reasons why you should always ask more questions when you hear a question. How to think like a detective. 
How data science is done in three different types of organizations. Three common mistakes people make when borrowing ideas. How we created our own agile methodology. The importance of finding your own answer.
The three types of errors in data science and how to deal with them. Why intelligent people make mistakes. How not to surprise yourself by errors you knew. The art of not making errors personal. The importance of thinking and talking trade-offs instead of errors. 
How data-intensive technologies have changed in the past five years, the best way for data scientists to stay on top of technologies, and the three timeless data roles.This episode is a guest interview with Wilco. Wilco has 20 years of experience in building tech, product teams, and big data architectures. He is the Chief Technology & Product Officer at ScaleForce, previously head of software engineering, head of product, and lead of innovation lab at trivago, as well as CTO and founder of venture-backed start-ups and scale-ups.Access 40 years of combined SaaS experience at: https://www.scaleforce.services/
When do you stop looking at the data, make a decision, and move on?  We dive deep into this audience question. But instead of giving an answer, we think that the best answers come from asking four more questions. We will show you what these questions are, why it makes sense to fight questions with questions, and how you can use them to unstuck your team and yourself. 
The number one pitfall of highly specialized roles, the consequence of premature optimization, the garden of many low hanging fruits, the hidden reasons why these giants publish more papers, and why you shouldn't blindly follow them. 
Three common mistakes about uncertainty in business, the idea of just enough uncertainty for decision making, pitfalls of p-value in AB testing, and how leaders can benefit from fostering conversations about uncertainty and data-driven decision-making in their organizations.
Why you don't need a perfect CV before applying, why you shouldn't try to answer all questions during interviews, the right mindset to think about hiring companies, and also some unsolicited relationship advice. Enjoy.
How to use the Puzzle Mapping technique to lead project kickoff meetings effectively, so that you can come up with concrete and feasible plans that everyone is happy about.  You can download an example Puzzle Map here. It is much easier to understand this technique when you see the example. Enjoy the episode. 
Why you should try five sub-optimal solutions instead of aiming for the optimal solution, why it is often better to write lower quality code at the beginning, and the importance of having discipline when you take shortcuts.  
How to evaluate new versus baseline when you already have an existing solution, how to use tracer bullets when there is no existing solution, and how to build accurate intuitions on both data science and business sides.  
Why it is important to avoid simplistic labels of maturity, how to measure competencies, the two natural ways to give feedback to data scientists, and the four key factors for creating development opportunities for your team.  
We got our first audience question about data scientist maturity model. To answer that, Nima and I talk about why it is important to avoid simplistic labels of maturity, how to measure competencies, the two natural ways to give feedback to data scientists, and the four key factors for creating development opportunities for your team. 
Min's journey from an individual contributor to a team lead, the importance of being explicit about uncertainty, how to get the most value out of offline evaluations, and other lessons she learned along the way. This is a guest interview episode with Min Fang. Min was trained as a computational linguist, worked as a data scientist, and became a team lead of data scientists and software engineers. She is interested in data-driven problem solving by applying natural language processing, machine learning, and statistical analyses. She also enjoys building strong teams that deliver these data-driven solutions. 
How to apply it to find common language between business and data science people, how to avoid the pitfall of shiny solutions, translating complex business needs to tangible requirements, and making your work more meaningful. 
Why you can't only rely on existing methods to evaluate your work, what to do when you don't have evaluation data, what to do when there is no ground truth, why some mistakes are much more important than others, and the importance of ongoing evaluations. 
Rate Podcast

Share This Podcast

Recommendation sent

Followers

1

Join Podchaser to...

  • Rate podcasts and episodes
  • Follow podcasts and creators
  • Create podcast and episode lists
  • & much more

Podcast Details

Created by
Naked Data Science
Podcast Status
Active
Started
Dec 4th, 2020
Latest Episode
Dec 4th, 2020
Release Period
Weekly
Episodes
28
Avg. Episode Length
25 minutes
Explicit
No
Order
Episodic
Language
English

Podcast Tags

Do you host or manage this podcast?
Claim and edit this page to your liking.
Are we missing an episode or update?
Use this to check the RSS feed immediately.