MLOps.community

Explore why data scientists should know data engineering with Dan Sullivan, a software architect and data scientist. Learn about the advantages, challenges, and transitions in AI, MLOps, and cloud platforms. Discover the intersections of data roles, data warehouses, and data lakes in efficient data processing. Enhance data science efficiency and modeling through iterative feedback and skills in data engineering.

Mercado Libre built Fury, a platform for machine-learning solutions supporting 500 users. They discuss platform features, technology, and Carlos de la Torre's mysterious LinkedIn denial. The podcast covers challenges in ML ops, expansion within Mercado Libre, and the evolution of machine learning practices in Latin America.

MLOps.community meetup #9 with Charles Martin - 10 years deploying Machine Learning in the Enterprise: The Inside Scoop!   Join the Community: ⁠⁠⁠⁠⁠⁠⁠⁠⁠https://go.mlops.community/YTJoinIn⁠⁠⁠⁠⁠⁠⁠⁠⁠Get the newsletter: ⁠⁠⁠⁠⁠⁠⁠⁠⁠https://go.mlops.community/YTNewsletterWhy do some machine learning projects succeed while others fall completely?   In this discussion, we will discuss the real-world challenges that Enterprises face in deploying ML solutions, focusing on challenges with existing, legacy dev-ops environments and how certain patterns of success emerge to help combat failure.    Dr. Martin runs a boutique consultancy in San Francisco, California, that supports organizations looking to research, build, and deploy data science, machine learning, and AI products.  He has worked with clients like eBay, BlackRock, and GoDaddy, as well as widely successful startups such as Aardvark (acquired by Google) and Demand Media (the first public billion-dollar IPO after Google). He is a world-renowned researcher, collaborating with UC Berkeley on the WeightWatcher project, and has taught at UC Berkeley and Stanford, and spoken at KDD, ICML, etc.   He is also currently a scientific advisor to the Page family’s Anthropocene Institute, consulting on areas including modern nuclear and quantum technologies and their response to the current pandemic.  Read more from Charles: http://calculatedcontent.com/ Join our Slack community: https://go.mlops.community/slack  Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/ Connect with Charles on LinkedIn: https://www.linkedin.com/in/charlesmartin14/

Meet up #10 Saurav Chakravorty sat down with us to talk about his vision of how MLOps reflects the old Indian story of blind men and an Elephant. Join the Community: ⁠⁠⁠⁠⁠⁠⁠⁠https://go.mlops.community/YTJoinIn⁠⁠⁠⁠⁠⁠⁠⁠Get the newsletter: ⁠⁠⁠⁠⁠⁠⁠⁠https://go.mlops.community/YTNewsletterAs a lead data scientist at Brillo, Saurav has built many MLOps pipelines and has experience using different ML platforms. He comes to talk with us about the difficulties of taking an ML platform from infancy to production and other key factors he has seen within the MLOps space.   Today, data science is a field that is an aggregation of people from various backgrounds - econometrics, statistics, engineering, business analysts, and data engineers. Each of these groups has different expectations from a Machine Learning platform. But, each group faces problems that have some common challenges - improving reproducibility, reducing technical debt, and reducing the time to try new experiments. The challenge before any MLOps system is to create platforms and processes that address the needs of each of these groups.   Saurav is a tinkerer in the Machine Learning world with experience in the design and development of ML applications and processes.  In the past few years, he has been focused on improving the processes and tools around the Machine Learning teams. He explores the ideas of Auto ML, ML Ops, and model evaluation. He helps customers adopt and use the best tools and processes that allow them to scale their Data Science or Machine Learning tools. He has development experience in the open stack ML platforms and, of late, the managed ML services from Azure and AWS.  You can read his article about creating your own MLOps pipeline with open source tools here: https://towardsdatascience.com/mlops-reducing-the-technical-debt-of-machine-learning-dac528ef39de  Join our MLOps community Slack:https://go.mlops.community/slackCome to our next MLOps meetup: https://tinyurl.com/yajmywre  Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/Connect with Saurav on LinkedIn: https://www.linkedin.com/in/sauravchakravorty/

Join the Community: ⁠⁠⁠⁠⁠⁠⁠https://go.mlops.community/YTJoinIn⁠⁠⁠⁠⁠⁠⁠Get the newsletter: ⁠⁠⁠⁠⁠⁠⁠https://go.mlops.community/YTNewsletterLinkedIn, Spotify, Volvo, JP Morgan, and many other market leaders are leveraging Kubeflow to simplify the creation and efficient deployment of Machine Learning models on Kubernetes.  This presentation will provide an update on the Kubeflow 1.0 release and review the Community’s best practices to support Critical User Journeys, which optimize ML workflows.As a data scientist will often need to build (and save) hundreds of variants of their model, this session will provide a deeper dive into how an integrated storage solution simplifies model-building and increases ML productivity.  The presentation will examine how to optimize the daily workflows of data scientists and eliminate complex and time-consuming manual tasks.   The talk will also highlight how efficient Kubeflow operations rely on Kubernetes storage primitives, such as Dynamic Volume Provisioning, Persistent Volumes, and StatefulSets.  This integrated solution simplifies the configuration, operations, and data protection for Kubeflow and generic K8S stateful apps in production-grade, multi-user environments.Bio:Josh Bottum is a Kubeflow Community Product Manager. His Community responsibilities include assisting users to quantify Kubeflow business value, developing critical user journeys (CUJs), triaging incoming user issues, prioritizing feature delivery, writing release announcements, and delivering Kubeflow presentations and demonstrations. Mr. Bottum is also a VP of Arrikto. Arrikto simplifies storage operations for stateful Kubernetes applications by enabling efficient local storage architectures with data durability and portability.  Arrikto is a core code contributor to Kubeflow.Join our MLOps community Slackhttps://go.mlops.community/slackConnect with Demetrios on LinkedIn: ⁠https://www.linkedin.com/in/dpbrinkm/Connect with Josh on LinkedIn:https://www.linkedin.com/in/joshbottum/

Join the Community: ⁠⁠⁠⁠⁠⁠https://go.mlops.community/YTJoinIn⁠⁠⁠⁠⁠⁠Get the newsletter: ⁠⁠⁠⁠⁠⁠https://go.mlops.community/YTNewsletterWhat does the MLOps pipeline at a London-based FinTech startup, TrueLayer, look like?  London-based fintech start-up TrueLayer decided to use Machine Learning instead of a rule-based system in mid-2019, and in our 7th meetup, we spoke to their lead data scientist, Alex Spanos, about everything that entailed.   During the meetup, we dove into how TrueLayer architected their MLOps pipeline for their Open Banking API: more specifically, which tools they use and why, what prompted them to use machine learning, and how Alex sees the role of a Machine Learning Engineer. Alex has led the hiring process of Machine Learning Engineers and shared learnings on candidates and businesses alike.   Alex is the Lead Data Scientist at TrueLayer, focusing on building Open Banking API products powered by data. Prior to TrueLayer, he built predictive models in Financial Services, used social data to predict the “next-big-thing” in fast-moving consumer Goods, and introduced Machine Learning techniques in subsurface imaging.  His academic background is in Applied Mathematics & Statistics.  Check out his blog entries for more info:https://blog.truelayer.com/improving-the-classification-of-your-transaction-data-with-machine-learning-c36d811e4257https://alexiospanos.com/hiring-machine-learning-engineers-part-1/https://alexiospanos.com/hiring-machine-learning-engineers-part-2/Connect with Demetrios on LinkedIn:  https://www.linkedin.com/in/dpbrinkm/ Connect with Alex on LinkedIn:  https://www.linkedin.com/in/alexspanos/Join us on Slack: https://go.mlops.community/slack

In our 6th meetup, we spoke with the CEO of Scribble Data Dr. Venkata Pingali.Join the Community: ⁠⁠⁠⁠⁠https://go.mlops.community/YTJoinIn⁠⁠⁠⁠⁠Get the newsletter: ⁠⁠⁠⁠⁠https://go.mlops.community/YTNewsletterScribble helps build and operate production feature engineering platforms for sub-fortune 1000 firms. The output of the platforms is consumed by data science and analytical teams. In this talk, we discuss how we understand the problem space, and the architecture of the platform that we built for preparing trusted model-ready datasets that are reproducible, auditable, and quality checked, and the lessons learned in the process. We will touch upon topics like classes of consumers, disciplined data transformation code, metadata and lineage, state management, and namespaces. This system and discussion complement work done on data science platforms such as Domino and Dotscience.Bio: Dr. Venkata Pingali is Co-Founder and CEO of Scribble Data, an ML Engineering company with offices in India and Canada. Scribble’s flagship enterprise product, Enrich, enables organizations to address 10x analytics/data science use cases through trusted production datasets. Before starting Scribble Data, Dr. Pingali was VP of Analytics at a data consulting firm and CEO of an energy analytics firm. He has a BTech from IIT Mumbai and a PhD from USC in Computer Science.Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/Connect with Venkata on LinkedIn: https://www.linkedin.com/in/pingali/Join us on Slack: https://go.mlops.community/slack

Join the Community: ⁠⁠⁠⁠https://go.mlops.community/YTJoinIn⁠⁠⁠⁠Get the newsletter: ⁠⁠⁠⁠https://go.mlops.community/YTNewsletterIn our 5th meetup, we spoke with the Brasilian ML Engineer Flavio Clesio.Machine Learning Systems play a huge role in several businesses, from the Banking industry to recommender systems in entertainment applications to health domains. The era of "A Data Scientist with a Script in a single machine" is officially over in high-stakes ML.We're entering an era of Machine Learning Operations (MLOps) where those critical applications that impact society and businesses need to be aware of aspects like active failures and latent conditions. This talk will discuss risk assessment in ML Systems from the perspective of reliability, safety, and especially causal aspects that can lead to the rise of silent risks in said systems.Slides for the talk can be found hereBio:Flavio Clesio is a Machine Learning Engineer (NLP, CV, Marketplace RecSys) and at the moment works at MyHammer AG, where he helps build Core Machine Learning applications to exploit revenue opportunities and automation in decision-making.Prior to MyHammer, Flavio was a Data Intelligence lead in the mobile industry and a business intelligence analyst in financial markets, specifically in Non-Performing Loans. He holds a master’s degree in computational intelligence applied in financial markets (exotic credit derivatives).This was a virtual fireside chat between Flavio Clesio, Demetrios Brinkmann, and the MLOps community. Relevant links can be found below. Join us on Slack: https://go.mlops.community/slackand register for the next meetup here.Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/Connect with Flavio Clesio on LinkedIn: https://www.linkedin.com/in/flavioclesio/

MLOps Community Meetup #4 With Shubhi JainJoin the Community: ⁠⁠⁠https://go.mlops.community/YTJoinIn⁠⁠⁠Get the newsletter: ⁠⁠⁠https://go.mlops.community/YTNewsletterIn the 4th online meetup for our MLOps.community We spoke with Shubhi Jain, Machine Learning Engineer, and an all-around great guy! Every organization is leveraging machine learning (ML) to provide increasing value to its customers and understand their business. You may have created models too. But, how do you scale this process now? In this case study, we looked at how to pinpoint inefficiencies in your ML data flow, how SurveyMonkey tackled this, and how to make your data more usable to accelerate ML model development.  Shubhi Jain is a machine learning engineer at SurveyMonkey, where he develops and implements machine learning systems for its products and teams. Occasionally, he’ll create YouTube videos about Machine Learning in collaboration with Springboard, an e-learning platform. He’s always excited to bring his expertise and passion for Data and AI systems to the rest of the industry. In his free time, Shubhi likes hiking with his dog and accelerating his hearing loss at live music shows.  This was a virtual fireside chat between Shubhi Jain, Demetrios Brinkmann, and the MLOps community. Relevant links can be found below. Join our MLOps Slack community: https://go.mlops.community/slackand register for the next meetup here.Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/ Connect with Shubhi Jain on LinkedIn: https://www.linkedin.com/in/shubhankarjain/Check out more of Shubhi on YouTube:https://www.youtube.com/watch?v=XsD2u7hAwI8https://www.youtube.com/watch?v=vcPNp21Mdg0https://www.youtube.com/watch?v=92kSljmHS7Uhttps://www.oreilly.com/strata-san-jose-2020/

MLOps community meetup #3! Last Wednesday, we talked to Phil Winder, CEO, Winder Research.Join the Community: https://go.mlops.community/YTJoinInGet the newsletter: https://go.mlops.community/YTNewsletter// AbstractPhil Winder of Winder Research joined us for the 3rd installment of our MLOps community meetup. In this clip taken from the long conversation, he speaks about why or why not he sees companies automating the retraining of Machine Learning Models. You can find the whole conversation here: https://www.youtube.com/watch?v=MRES5IxVnME.The topic of conversation for our virtual meetup was an in-depth look at a pyramid of software engineering best practices that built up to incorporate data science best practices. That is to say, we analyzed “the essentials”, "nice to have," and "optimal" ways of doing data science. Machine Learning/Data Science/AI is an extension of the technical stack. So you can't really talk about Data science best practices without accidentally talking about software engineering best practices. For example, model provenance doesn't count for anything if you don't have code or container provenance.  Just as Maslow has the basic human needs, so too do we have basic MLOps needs. Where does "MLOps", as a "thing", start and end? For example, the four very reasonable best practices of the operation of models, but these are usually consumed into higher-level abstractions because there is a lot more to do than "just" provenance.  // BioDr Phil Winder is a multidisciplinary software engineer and data scientist. As the CEO of Winder Research, a Cloud-Native data science consultancy, he helps startups and enterprises improve their data-based processes, platforms, and products. Phil specializes in implementing production-grade cloud-native machine learning and was an early champion of the MLOps movement. More recently, Phil has authored a book on Reinforcement Learning (RL) (https://rl-book.com), which provides an in-depth introduction to industrial RL to engineers.  He has thrilled thousands of engineers with his data science training courses in public, private, and on the O’Reilly online learning platform. Phil’s courses focus on using data science in industry and cover a wide range of hot yet practical topics, from cleaning data to deep reinforcement learning. He is a regular speaker and is active in the data science community.  Phil holds a PhD and M.Eng. in electronic engineering from the University of Hull and lives in Yorkshire, U.K., with his brewing equipment and family.// This was a virtual fireside chat between Phil Winder and Demetrios Brinkmann. The relevant links can be found below:Join our MLOps Slack community: https://go.mlops.community/slack  Connect with Demetrios on LinkedIn: https://www.linkedin.com/in/dpbrinkm/  Connect with Phil on LinkedIn: https://www.linkedin.com/in/drphilwinder/Follow Phil on Twitter: https://twitter.com/DrPhilWinder Learn more about Phil's company, Winder Research: https://winderresearch.com/