We are bringing you new DSB experience – podcast. Audio is in Czech language and we will try to accompany every podcast epizode with insights in english – mostly AI generated from the original content (thanks to Whisper, LLMs and more).
Feedback welcomed to hosts directly or to podcast@datasciencebulletin.com.
Links/Sources mentioned:
We went through few of the 39 lessons on Building ML Systems, Scaling, Execution, and More. We discussed points 1, 7, 3, 5, 8, 9, so a lot still remains for dear reader to explore.
Ben Affleck and his take on AI impact on movie industry.
Jakub made a prediction, that the most skilled individuals will relly on human skills not the AI ones. I will remind him in a year 😉
Perplexity.AI response on different data roles:
- Data Engineer: Focuses on building and maintaining data infrastructure, ensuring data is accessible and optimized for analysis. They work with technologies like SQL and big data tools (e.g., Hadoop, Spark)
- Data Analyst: Analyzes datasets to extract actionable insights, often using statistical tools and visualization software. Their goal is to support business decisions by interpreting data trends
- Data Scientist: Combines skills from both roles, employing advanced analytics and machine learning to develop predictive models and derive deeper insights from data
- Machine Learning Engineer: Specializes in deploying machine learning models, focusing on coding and production systems rather than data analysis
- Machine Learning Researcher: Engages in developing new algorithms and advancing machine learning methodologies, often in academic or R&D settings
- AI Engineer: Concentrates on creating AI systems that can learn and adapt, leveraging both machine learning and software engineering skills
SWYX and his 1.5 year old post The Rise of the AI engineer
Generated insights from the epizode
Welcome to the first-ever episode of the DSB Podcast! Hosted by Martin Münch and Jakub Kramata, two data scientists from ČSOB, this podcast is designed to bring hands-on perspectives from the world of data science and machine learning.
Why a Podcast?
The DSB—short for Data Science Bulletin—is a natural evolution of an internal ČSOB newsletter we’ve been running for 7 years. We originally started the newsletter to share news from the world of data science. Over time, we realized that we wanted a more opinionated format where we could dive deeper into topics that matter to us. While there are already plenty of podcasts with interviews and broad overviews, we felt there wasn’t much out there (especially in Czech) discussing the nitty-gritty of data science, machine learning, AI, …
Building Effective Machine Learning Systems
One of the big topics in this episode was lessons learned from building effective machine learning systems, inspired by Eugene Yan’s famous “39 Lessons on Building ML Systems, Scaling, Execution, and More.”
The real world is messy. Data is chaotic, unpredictable, and full of surprises. If someone tells you that AI is a walk in the park, they’re either lying or haven’t dealt with real-world systems.
You don’t always need machine learning. It’s valuable to start without machine learning, gives a data scientist insight into data, generates quick baseline to overcome.
Evals are a differentiator and moat. Having eval datasets that enables quick iteration is the biggest advantage you can have now.
Design with the data flywheel in mind. Besides evals, feedback loops that company can elevate quickly is the advantage to aim for.
Breaking Down the Roles in Data Science
There are many roles in data space. See the Perplexity.AI description above. Honorable mention is Data Miner – role no longer with us, but still important and useful.
Data engineering is the single most important step for any organization wanting to start with data. Analyst second – knowing what is in the data is more important than what data science can bring you.
Data analyst ability to talk with business counter parties is a overlap with data scientist and every DS should be able to do it.
Be First to Comment