Shreya Shankar
PhD student in databases focusing on MLOps @Berkeley_EECS @UCBEPIC. Prev. 1st ML engineer @viaduct_ai, research @googlebrain, BS & MS @Stanford CS. She/they.
View on 𝕏Threads
@ClementDelangue I don’t think tiered product offerings are bad if the business value difference is clear and quantifiable. Depends on your paying customers — if they’re technical,...
I’ve noticed some companies offering “tiered” versions of their ML API products. It makes sense to me from a training/inference cost perspective but no sense to me from a customer...
I’ve been frustrated for a while about the lack of diversity in engineering and data science roles at early-stage startups. So I’m starting a small mentorship circle for women and...
@seanjtaylor I think about this two ways: * what are the eng tools we need to facilitate multiple people working on ML for the same prod solution? CI/CD, @MLflow model promotion,...
I have been thinking a lot about designing systems for reproducibility in ML experiments from a lens of identifying the right pain points, realistic solutions, and good UX.
practical MLE tip: if you know your distribution isn’t Gaussian, min-max normalize instead of standardize https://t.co/zP6ivOFy7d
i love this thought experiment. i played piano & violin growing up. i dreaded Hanon & Rode exercises. i wondered why i had to learn boring pieces from different time periods. but l...
Recently a GPT-3 bot said scary things on Reddit and got taken down. Details by @pbwinston: https://t.co/idIWy1XEzj These situations create fear around "software 2.0" & AI. If we...
grafana & kibana remind me of these twin bullies in my elementary school. i could never remember which did what. they said a lot of things & i usually didn’t know how to respond. b...
In good software practices, you version code. Use Git. Track changes. Code in master is ground truth. In ML, code alone isn't ground truth. I can run the same SQL query today and...
i preach “iterate on the input data, not the model” a lot, but i want to add a recent reflection: *not all raw data needs to be featurized and fed to a model*
every morning i wake up with more and more conviction that applied machine learning is turning into enterprise saas. i’m not sure if this is what we want (1/9)