Feature Engineering Bookcamp by Sinan Ozdemir.
Sinan holds both a BA and MA in Pure Mathematics from Johns Hopkins University. His graduate work focused on algebraic geometry with applications to cybersecurity. He was Founder and CTO of Kylie.ai and Legion Analytics. Today Sinan is the Founder and CTO of LoopGenius. He is also a former Lecturer/Adjunct Professor, teaching graduate-level Business Analytics, Mathematics, and Computer Science.
Sinan is addressing a very critical, yet often overlooked stage of machine learning pipelines. A transformation of raw data into informative features will make or break your efforts. He is advocating the quality of the input data is the true measurement for any model’s performance.
He is presenting to readers six hands-on projects to upgrade your training data using feature engineering. This “Bookcamp” thereby prioritizes project-based curriculum over theory since structures the text is the true craft for data scientists. Instead of focusing on mathematical transformations, Sinan asks the reader to solve business problems, such as predicting flight delays.
LLMs need high quality data
By learning code-driven case studies, readers will have insights to social media classification, COVID detection, recidivism prediction, and even stock price movement detection. These lessons can result in key improvements to any organization’s machine learning pipelines. Theses ideas prove how a poorly chosen transformation introduces data leakage resulting in models gaining data outside of real-world scenarios. This also saves time and money fine-tuning parameters. This is in fact, clearly written for experienced machine learning engineers deploying with Python. A basic understanding of Pandas and Scikit-Learn is also required,
In conclusion, Feature Engineering Bookcamp is really key for machine learning engineers looking to build robust, production-ready LLMs. In the era of GenAI, Sinan reveals the ability to engineer high-quality data remains the key valuable skill.