Ahead of his presentation at the Machine Learning Innovation Summit in New York on December 11 & 12, we spoke to Haftan Eckholdt, Chief Data Science Officer at Plated.
Haftan Eckholdt is Chief Data Science Officer at Plated, a meal kit company headquartered in Manhattan. His early career included academic research appointments in Neuroscience followed by industry research appointments at companies like Amazon and AIG. He holds graduate degrees in Biostatistics and Developmental Psychology from Columbia and Cornell Universities. In his spare time he thinks about things like chess, cooking, cross country skiing, jogging, and reading. When things get really, really busy, he actually plays chess, cooks delicious meals and jogs a lot. Born and raised in Baltimore, Haftan has been a resident of Brooklyn since 1990.
Why do you think we have seen machine learning use increase so dramatically in the past 3 years?
Twenty three years ago: hard disk prices were dropping, so data starts accumulating (data were literally thrown on the floor on NYSE every weekday from 9:30am to 4pm, for real, you can look it up on the internet); ram prices were dropping, so data on the disk was in memory for calculations (equations could have more than twenty variables, if only you could code you own GBM from scratch), and R was invented, so long equations could be had by all. Twenty years later (three years ago) Yann LeCun began producing graduates in Data Science at NYU. Thusly, the uptick in machine learning.
How do you think organizations could be utilizing machine learning better?
Machine learning needs a cool dry host environment, and a designated role to play in decision making. Most organizations do not coordinate those two things with modeling, so that ten models are built for every one model hosted, and ten models are hosted for every one model used.
What are the biggest challenges currently facing the further spread of machine learning?
Operationalization. Most business practitioners have never met a modeler, much less used a model. No models can thrive until people know what modelers can do, and how to use models.
Do you think that machine learning regulation is currently fit for purpose?
Models have been regulated for many many many years in some industries (think AIG, NASA, etc.), where individual models have risk managers. So, where it really matters, we regulate. Do you want to pay for a model risk manager to oversee the selection of what movie to watch tonight <insert child/sibling/spouse/parent reference here>? Probably not.
What can the audience expect to take away from your presentation in New York?
Data Science is optimized for success at Plated, and this is a very rare thing. I want people to understand why I think this is so in terms of function, capability, and responsibility with some case studies. People new to the field should come away with a list of things to look for in their next role (what to ask in your next interview), and seasoned researchers should see what it can look like when you have what you need (what to put in your next budget request).
You can hear Haftan's presentation at the Machine Learning Innovation Summit in New York on December 11 & 12