Traditional Feature Engineering as A Still Important Pillar, Alongside Embeddings & Prompts
Plus: Announcing that 3 months of blood, sweat, & tears has come to fruition!
Hey everyone!
So for the first 3-4 months of joining Featureform, this 30+ page guide on “traditional” feature engineering was my constant companion.
Link: http://qr.codes/pBSMCL
Starting from an initial 500 word article that had been written by a contractor, I decided that what data scientists really needed (especially ones that were just getting into the field) was a thorough treatment of what feature engineering looks like, especially in production.
Of course, this was right before the prolific use of ChatGPT for writing blog posts really took off, so everything was manually reviewed, researched, written, & drawn before our previous designer was able to get at it.
Does traditional feature engineering still have a place?
I think so. At the very least, that’s the bet I’m making.
Please swing by and check out the guide, would make me feel free great on my birthday. :D
Some of my favorite diagrams include:
You've done such a remarkable job incorporating every single detail into this amazing ML lifecycle project guide. Kudos to you! 🤗