MLOps Awards 2022
For the first time ever we will be acknowledging the incredible work that has been put into the MLOps ecosystem over the past year.
These MLOps Industry Awards are of the highest and utmost honor. Receiving one will forever immortalize its owners in the hall of MLOps fame. Hollywood has the Oscars, the music industry has the Grammys and Machine Learning in production has an MLOp!
Without further ado, let’s get into the most prestigious (and only) awards in the production ML ecosystem.
Best Blog Post
Instacast had an incredible year of blogging. They showed the rest of us how advanced they really are behind the scenes. When this blog was released it was reshared hundreds of times by just about everyone in the MLOpsphere.
Let’s face it, real-time ML is on everyone’s mind these days. Maybe because it’s an absolute pain to get up and running, perhaps because we think it’s an absolute must for ML. Maybe it’s the logical next step in our evolution.
Whatever the case, this blog post struck a nerve and has since become the monumental real-time post of 2022.
Congrats to Guanghua Shu and Sahil Khanna on the win.
Serve Hundreds to Thousands of ML Models – Ernest Chan
Best MLOps Book
Reliable machine learning – Cathy Chen, Niall Richard Murphy, Kranti Parisa, D. Sculley, Todd Underwood
Many of us have been patiently waiting for this book to finally be released in its entirety. Now that it has been we can safely say it was worth the wait. With the all-star line-up of authors, how could it have been anything less than amazing?
For those looking to gain deep insights into Production ML, search no further.
Designing Machine Learning Systems – Chip Huyen
Eduardo Bonet – Gatekeeping ML
The first ever slack thread to get over 100 replies. This one got heated, went on tangents, and got some serious attention.
How did it all start off?
I hate the gatekeeping with the goal of thought leadership that “ml production code” has become across every single community. The reality is that you are not even a software engineer is paid to write good code, we are paid to deliver results (and writing good is a way to facilitate that). Yet, I see folks creating this “aasolute do’s vs absolute dont’s” like inly they hold the truth entire experience of the field banking on the naiveness of new users who can’t yet understand how much bullshit that is.
The reality is. Everything single thing I said that must or mustn’t be done so far, I eventually found a situation where doing the opposite was the most appropriate choice.– Eduardo Bonet
Most Cringe Thread
Data Science Habits – Will Angel
I know not all data scientists are like this. Sadly, a few of them are giving the discipline a bad name.
Case and point? This observation by Will Angel.
I just watched a data scientist copy and paste code from a notebook into vscode, save the file to disk, and email it to us so we can put it into production while he’s on vacation in Hawaii.
Is this recoverable or do I need to find a new data engineering job?– Will Angel
Let’s face it, if it wasn’t true it wouldn’t hurt so much.
Best Talk Given at a Conference
Daniele Perito – Empowering Small Businesses with the Power of Tech, Data, and ML
This talk given at the apply(conf) in February has just about everything you could ever want from a presentation. Current ML stack, the design decisions that went into creating it, tradeoffs they had to deal with and retro learnings.
I especially appreciate how honest he was about where they need to improve and what the next steps are in maturing the ML stack.
Operationalizing Machine Learning: An Interview Study – Shreya Shankar, Rolando Garcia, Joseph M. Hellerstein, Aditya G. Parameswaran
Why was it so good? Non-Biased interviews of MLEs that came to a conclusion which was not “buy xyz tool to alleviate this pain”. Many of the engineers interviewed for this study came from the MLOps community itself, which obviously gives the paper some serious bonus points!
Machine Learning Operations (MLOps): Overview, Definition, and Architecture paper – Dominik Kreuzberger, Niklas Kühl, Sebastian Hirschl
Best Podcast Episode
Investing in MLOps – Leigh Marie Braswell and Davis Treybig
With so many incredible episodes this year this may have been the most cutthroat awards category. This was the closest award decision and the judges barely etched out a winner. In the end, this conversation between two great investors deep in the ML space won over the judges.
Billions have been invested into the dev tooling for MLE category over the past year. This conversation walks through the rationale behind that and how some of the decisions are panning out. Davis and Leigh Marie give objective opinions on the current state of the MLOps ecosystem and where we are going next.
Most Mentioned Tool
Maybe because it’s such a pain. Maybe because AWS support is pretty much non existent. Maybe because they have the lion share of the market.
This tool was referenced at least once a week in some form or fashion.
Most Underrated Resource
Most of us know about the great engineering blogs that Uber, AirBnB, and Netflix have. But how many of us take note of the consistently high-quality blog posts coming out of a random Brazilian challenger bank?
If this one wasn’t in your bookmark bar before, you may want to add it.
Best Vendor Feud
Snowflake and Databricks
Next year they will both be holding their yearly conferences on the same dates making us all have to choose between one or the other. Also how many posts did you see this year about how data lake is a better architecture than a data warehouse? Or some benchmarking post about one of these being faster than the other.
Grab your popcorn and tweet about how much better one of these tools is than the other.
Best MLOps Swag Photo
Nuff said. K8s is a gateway drug
About halfway through the year my LinkedIn feed became dominated by clear and concise diagram designs that made me stop scrolling every time. Aurimas was able to intuit exactly what the MLOps world needed and put it out there for the world to see. I have been learning from him and am excited to see what he does in 2023 with all this newfound clout!
Best Marketing (by a tooling company)
I heard they had 50k registrants for the event. There was just the right blend of visionaries and practitioners over the 3-day event. The theme of the event was “operationalizing ML at scale” aka MLOps, although I didn’t completely agree with that (it felt like they were just grabbing at buzzwords) you can fault them for trying.
The execution of this event was well played. To the Scale AI marketing team, we salute you.
So there you have it, folks. the first annual MLOp Awards has come to an end. If you feel there is some category we missed, please write us so that we can include it in next year’s ceremony.