Summary of The Data Janitor returns | Daniel Molnar

00:00:00 - 00:45:00

In this video, Daniel Molnar discusses the data janitor and why it is important to have a good understanding of data science in order to be successful. He also provides advice on how to achieve this, emphasizing the importance of having a good sample size.

• 00:00:00 Daniel Molnar gives a presentation on the state of data in the industry. He discusses the various stages of data gathering and how data management has changed in recent years. He also discusses how companies are trying to get more data logging into systems, and how machine learning and deep learning are affecting the industry. He ends the presentation by giving a pyramid of data needs for a company, and how the different roles in a company interact.
• 00:05:00 In this YouTube video, Daniel Molnar discusses the importance of data labeling and provides an example of how to make a programmatic KPI definition. He also warns about the dangers of relying too heavily on Google Analytics data.
• 00:10:00 Daniel Molnar discusses the data janitor and why it is important to have a good understanding of data science in order to be successful. He also provides advice on how to achieve this, emphasizing the importance of having a good sample size.
• 00:15:00 The Data Janitor returns to talk about the main problems with data engineering and how to solve them. He recommends looking to Viki Bark for advice on the best way to approach data engineering.
• 00:20:00 The Data Janitor returns to discuss deep learning, which is a field of computer science that uses artificial intelligence to learn patterns in data. He talks about three recent papers that demonstrate the inefficiency of deep learning models, and how this could lead to the future use of GPUs in lieu of CPUs for certain tasks. Finally, he talks about a recent paper that suggests that deep learning models can be constructed using less efficient methods, which could lead to more efficient models in the future.
• 00:25:00 The Data Janitor discusses how his Python script became 10 times faster and consumed 1/10 of the resources, resulting in two magnitudes of improvement. He mentions that data engineering is a skill that is useful in other fields, and recommends watching Jordan Blows's recent presentation on preventing the collapse of civilization.
• 00:30:00 The Data Janitor video looks at how migrating data from one cloud platform to another can be difficult and time-consuming. The Janitor also compares the performance of different cloud platforms, including AWS, GCP, and Azure. The main takeaway is that HDFS is always faster than S3, and that spark can be upgraded to become faster over time.
• 00:35:00 The Data Janitor returns to discuss Facebook's new Presto data warehousing solution. Daniel reviews the advantages and disadvantages of Presto against other options, and discusses the potential for front-end tracking issues. He also recommends using Google Analytics alternatives if you're unable to use its tracking features.
• 00:40:00 The Data Janitor tool helps developers debug and troubleshoot data issues. The video's main subject is how to employ the Data Janitor to help with data engineering tasks and to prevent data loss.
• 00:45:00 Daniel Molnar discusses CRM and how it can be improved. He notes that while CRM is important, it can be improved by ensuring that the data is trustworthy and double checking it. His final message is that it is important to be proactive in improving marketing processes, even if it means changing business rules or behavior.