Oleg Avdeev

@dialtone_ @NextRoll_Inc Congrats VV! Best of luck with the next thing!

Well... it's been 15 years that I've started working on @NextRoll_Inc and I've made the hard decision to take on a new challenge. I've matured and learned so much while working on that product and with one of the best engineering teams out there. Heartfelt thanks folks.

We're about to start out full stack recommender system tutorial live code along with @jacopotagliabue for @OuterboundsHQ. come say hi! youtube.com/live/PkKs6oWRE…

I’m really excited about the Metaflow as a service offering by @OuterboundsHQ Developer ergonomics are among the best of any tool I’ve seen comparable to Keras outerbounds.com/blog/announcin…

@msjpenn @zackkanter Exactly, "no machine learning, no AI" my ass. Whoever wrote that paragraph needs to learn basic math before even using those words in a sentence.

@zackkanter Lol, I think Twitter has been a dumpster 🔥 of wasted engineering, but this is 9000 tweets manually investigated / quarter. Original Musk tweet was 100 total for sample, now it’s “100 per day.” Random sample size of 9000 is plenty to confidently say < 5% is spam

People far overestimate how well most tech companies are run outside of a few shining exceptions. Twitter has long been three kids in a trench coat pretending to be a serious organization, let alone in areas where there’s an incentive to lack rigor.

Infrastructure engineers ❤️ @kubernetesio for many valid reasons. Data scientists shouldn't have to care about it. They deserve a user experience optimized for their needs, not for the needs of infrastructure. We built a bridge between the two worlds 🌉 🎉

@adamlaiacano The key was to be lucky with the first letter Wordle 304 2/6 🟩⬜⬜🟨🟨 🟩🟩🟩🟩🟩

Ok everyone. Let’s see your L’s today. Wordle 304 X/6* ⬛🟨⬛⬛⬛ 🟨🟩⬛⬛⬛ ⬛🟩⬛🟩🟩 ⬛🟩⬛🟩🟩 ⬛🟩⬛🟩🟩 ⬛🟩⬛🟩🟩

If you'd been wondering where to donate to help the Ukrainians, this is it. Ryan runs a logistics startup, and his humanitarian efforts are startup-level efficient. You can be confident that he'll ship the right stuff, quickly, and that it will end up in the right hands.

It is such a privilege to get to work with fun, wise, and empathetic data scientists and Pythonistas like @hugobowne 🙇 Would you like to join us? outerbounds.com/workwithus/

📢 THE DAY HAS FINALLY COME Announcing 🥁 🔥🔥Notebooks in Production With Metaflow.🔥🔥 Orchestrate notebook execution in your DAGs & visualize/debug production workflows. Outputs are rendered in the MetaFlow GUI 🚀 outerbounds.com/blog/notebooks… But that's not all! 🧵 👇 1/n

@bernhardsson In the Reed Hastings book IIRC he says they spend substantially more on travel as a result of this, so I guess it is not as clear cut from pure financial POV (ofc the argument is that it is still worth it for cultural reasons)

Is there any good reason not to adopt a Netflix style "do what's in the company's best interest" type expense policy? I suspect (time wasted on expense policy) >> (money lost from expense policy abuse) for almost all companies

I'm writing a book about #datascience and #machinelearning infrastructure based on my experiences at Netflix and earlier companies. The book uses #metaflow to demonstrate the concepts but it is not exclusive to Metaflow. I'd love to hear your thoughts and feedback!

"As an informal estimate, our #datascience team believes they were able to test twice as many models in Q1 2021 as they did in all of 2020, with simple experiments that would have taken a week now taking half a day." - CNN 🤯 medium.com/cnn-digital/ac… #MachineLearning #MLOps

fun: the "Stealth Startup" org on LI is showing an impressive 6x headcount growth over the last two years:

Neither @JackieFielder_ nor @Scott_Wiener are talking about a SERIOUS plan to flatten the geography of SF. Our plan ELIMINATES height discrepancies of over 10 meters AND creates jobs. #winwin #paysforitself #flattenSF

Great podcast with our CTO @kevinmstumpf on Software Engineering Daily discussing ML platforms at Uber and Tecton. softwareengineeringdaily.com/2020/06/03/tec… #machinelearning #mlops #DataScience pic.twitter.com/hCe0MSW89U

everyone: Lisp is a functional programming language, deeply inspired by rich mathematical history John McCarthy:

@dmitriid @awscloud Having built a version of this (using modern web technologies!) that only uses 1MB of resources instaguide.io , tbh I don't feel it makes all that much difference usability wise

Modern web is cancer. @awscloud pricing page which is a static page with *just text, no images* is - 179 requests - 34 MB of resources - 42 seconds load time They need all that to do an "onChange=function(){ table1.hide(); table2.show(); }"

We use AWS Batch a lot, like really a lot. tech.nextroll.com/blog/dev/2019/… Here's how we do it and what we learned about it and even a few tools we built and open-sourced while doing it all. @mndoci can we get more capacity now? :)

Here's a recording of my talk "More Data Science with Less Engineering: ML Infrastructure at Netflix" from SF Big Analytics Meetup at Lyft youtube.com/watch?v=FJ2lGf…