Making MLOps Marvelous with Maria Vechtomova

About this episode

What does it actually take to move machine learning from experiments into production reliably, responsibly, and at scale?

In this episode of Alexa’s Input (AI), Alexa talks with Maria Vechtomova, co-founder of Marvelous MLOps and an O’Reilly author-in-progress on MLOps with Databricks. Maria shares how her background in data science led her into MLOps, and why most teams struggle not because of tools, but because of missing processes, traceability, and shared understanding across teams.

Alexa and Maria dive into what separates good MLOps from fragile deployments, why shipping notebooks as “production” creates long-term pain, and how traceability across code, data, and environment forms the foundation for reliable ML systems. They also explore how LLM applications are reshaping MLOps tooling, and where the biggest skill gaps still exist between platform, data, and AI engineers.

A must-listen for anyone building, operating, or scaling machine learning systems and for teams trying to make MLOps less magical and more marvelous.

Learn more about Marvelous MLOps and Maria’s work below.

Links

Watch: ⁠⁠https://www.youtube.com/@alexa_griffith⁠⁠

Read: ⁠⁠⁠⁠https://alexasinput.substack.com/⁠⁠⁠⁠

Listen: https://creators.spotify.com/pod/profile/alexagriffith/

More: ⁠⁠https://linktr.ee/alexagriffith⁠⁠

Website: ⁠⁠https://alexagriffith.com/⁠⁠

LinkedIn: ⁠⁠https://www.linkedin.com/in/alexa-griffith/⁠⁠

Find out more about the guest at:

LinkedIn: https://www.linkedin.com/in/maria-vechtomova/

Takeaways

Maria started as a data analyst and transitioned into MLOps.

She emphasizes the importance of tracking data, code, and environment in MLOps.

MLOps is a practice to bring machine learning models to production reliably.

Good deployment processes require modular code and proper tracking.

MLOps differs from DevOps due to the complexities of data and model drift.

Education is crucial for bridging gaps between teams in AI.

Small steps can lead to better MLOps practices.

Scaling MLOps requires understanding the unique data of different brands.

The rise of LLMs is changing the MLOps landscape.

Effective teaching methods involve step-by-step guidance.

Chapters

00:00 Introduction to MLOps and Maria's Journey

02:11 Maria's Path to MLOps and Knowledge Sharing

04:41 The Importance of MLOps in AI Deployments

10:12 Defining MLOps and Its Challenges

11:38 MLOps vs. DevOps: Key Differences

13:00 Overcoming Stagnation in MLOps

16:04 Small Steps Towards Better MLOps Practices

19:29 Scaling MLOps in Large Organizations

21:58 The Impact of LLMs on MLOps

23:58 The Shift from Traditional ML to AI Applications

26:51 Evolving Roles in AI Engineering

28:33 Databricks: A Comprehensive AI Platform

31:45 Future of AI Platforms and Regulations

34:26 Bridging Skill Gaps in AI Teams

38:42 The Importance of Context in AI Development

40:40 Foundational Skills for MLOps Professionals

45:43 Integrating Personal Passions with Professional Growth

47:30 Building Impactful AI Communities