Maria Vechtomova

Senior Machine Learning Engineer

Ahold Delhaize

Working on improving MLOps maturity across the brands of Ahold Delhaize by creating methodology, standards and tools that make ML deployment faster, more secure and transparent.

Manager ML Engineering | MLOps Tech Lead

Marvelous MLOps

Co-Founder

Migrating machine learning pipelines from private cloud to AWS, building new machine learning pipelines using AWS Python CDK.

Creating new Analytics tools and infrastructure that is used by Analytics teams at KPN, looking for new opportunities to use data science techniques at KPN, experimenting with new packages/ models/ infrastructure.

Data Scientist | Machine Learning Engineer

Using statistical and machine learning techniques in order to extract value from the data, consulting internal business stakeholders on possible business opportunities based on the data, combining various data sources and building the data flows in order to create business dashboards.

Data Analyst | Data Scientist

Udacity

Mentor and code reviewer

I have been in the Data & AI field for over a decade, and have dedicated the largest part of my career to MLOps - something I am truly passionate about.

MLOps is not about the tools but about the principles. However, things become easier with the right tools. I know that as I have seen many tools and have built tools myself.

I have worked with Databricks for over 3 years, and I believe that it is one of the best platforms for MLOps nowadays. However, there is too little information on how to do it properly.

My mission is to educate data & AI professionals on MLOps, and I leverage Databricks to explain how to follow MLOps principles. Currently writing a book for O'Reilly and teaching about MLOps with Databricks on Maven and LinkedIn Learning.

I’ve been pessimistic about Colab for years.

Not because it wasn’t useful, it absolutely was. Because it made real software engineering practices way harder than they needed to be. 

But… things change. Google dropped something I genuinely didn’t think we’d see this soon: the official Google Colab extension for VS Code.

This is a big deal! Finally promoting modern development workflows instead of living in “classroom demo” land. Go install it!

Thanks to Chris "The Wiz 🪄" Alexiuk for sharing it first.

MLOps Tech Lead | Marvelous MLOps | Follow me to learn about MLOps | Public speaker

AI @ NVIDIA - 🏗️ Build | 🚢 Ship | 🚀 Share

Co-founder @Cauchy | Global speaker & advisor | MLOps Tech Lead | Writing a book for O’Reilly

Daniel Jimenez

A year of change, challenge, and growth.

After many years in a corporate environment, I knew I wanted to start something of my own… I just didn’t know what that something was. So I began exploring.

Alongside my full-time job at Ahold Delhaize, where I’m now entering my final couple of months, I:

- started building a community here on LinkedIn, and realized I love it
- taught four cohorts of the MLOps course on Maven, and realized I love it
- have been writing a book (it’s so close to the finish line!)

It’s been intense. Many late nights (and yes, Marx has been by my side through all of them. Not always comfortable… he does not care about ergonomics 😅).

But through all of this, I discovered something important:
💎 What truly drives me is helping people change the way they work and think, for the better.

For the longest time, I said: “I do not want to be a consultant.” Funny how life works: that’s exactly where I’m headed.

But not to “build solutions and walk away.” I want to leave a mark and empower people to build great AI products - not make them dependent.

That’s the idea behind Cauchy, named after French mathematician, whose work laid the foundation for gradient descent. Without that, modern AI wouldn’t exist.

If you’re curious to chat about how we can help your team succeed with AI:

➡️ Meet me, my co-founder Casper Lubbers, and founding partner Eirini Papakosta at the Databricks Data+AI World Tour in Amsterdam
➡️ Comment to the post/ send a DM, and I will reach out to you 
➡️ Send us message: https://lnkd.in/e3uYGKEm

MLOps Tech Lead | Writing a book for O'Reilly | Databricks MVP

As an ML/AI engineer, I depend heavily on infrastructure, but actually deploying that infrastructure is rarely the fun part.

I’ve fought with Terraform enough to know it’s not for me: clunky, slow to iterate, and painful to debug. I prefer Python-first tools like the AWS CDK, but that doesn’t help when I need the same experience on Azure.

Pulumi has been on my “should really try this” list for a while, and when Pulumi released Pulumi Neo, I finally dove in.

Neo gives you the Pulumi experience with basically zero effort. It’s a chat interface inside Pulumi: describe what you want, and it builds + deploys the infra.

I deployed a DynamoDB table without writing a single line of code. My workflow:
- create an IAM user
- add credentials to Pulumi, share account + region
- create a Pulumi GitHub app
- create a Pulumi secret and store it in my GitHub repo
- tell Neo: “spin up a dynamodb table”

That’s it. Fewer iterations than I expected, and I ended up with a repo containing working infra code and deployment pipelines.

And sure, this was a toy example. Neo can go way bigger: updating Kubernetes clusters, managing complex configs, deploying full multi-service setups.

Pulumi + Pulumi Neo is absolutely worth checking out.

Try Pulumi Neo (it is free, no credit card required!): https://lnkd.in/eRenmyin
Learn more about Pulumi Neo: https://lnkd.in/ex7zuSiX

--------------------------------------------------------
I partnered with Pulumi for this post, but everything here reflects my own experience.

Pulumi

A week ago, I shared a post about unexpected costs when migrating from legacy online tables to the new ones that use Lakebase.

Impressed by how fast Databricks responded. The docs have now been updated to make the cost implications and best practices much clearer for end users.

Cost optimization best practices 👇 :
➡️ Reuse online stores: You can publish multiple feature tables to a single online store. For development, testing, or training scenarios, consider sharing one online store across projects instead of creating separate ones.
➡️ Right-size capacity: start with the smallest instance for testing and only scale up if you truly need more performance.
➡️ Delete unused online stores. Online stores incur continuous cost:  if you're not using one anymore, delete it.

Kudos to Denny Lee and the team for jumping on this so quickly.

PM Director, Developer Relations at Databricks | Apache Spark™ & MLflow Contributor | Delta Lake & Unity Catalog Maintainer | Co-host of Data Brew Podcast | Advisor to Early-Stage AI & Data Startups

taplio

Maria Vechtomova's Linkedin Analytics