> Markdown version of [/videos/369-the-state-of-mlops-machine-learning-in-production-at-enterprise-scale?t=1343](https://www.wearedevelopers.com/videos/369-the-state-of-mlops-machine-learning-in-production-at-enterprise-scale?t=1343). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md).

---

# The state of MLOps - machine learning in production at enterprise scale

Successful MLOps depends less on vendor tools and more on rigid software engineering. Learn how to transition from isolated notebooks to robust, cross-functional ML pipelines at enterprise scale.

- **Speakers:** Bas Geerdink
- **Event:** WeAreDevelopers LIVE
- **Published:** March 25, 2022
- **Duration:** 47:19
- **URL:** https://www.wearedevelopers.com/videos/369-the-state-of-mlops-machine-learning-in-production-at-enterprise-scale

## Summary

Transitioning machine learning from experimental notebooks to robust production environments demands a fundamental shift in both team culture and software architecture. Treating ML models as formal software artifacts requires leaving isolated Jupyter environments behind and adopting dedicated MLOps methodologies. By merging data scientists, developers, and operations engineers into cross-functional units—much like the DevOps revolution—organizations can accelerate deployment speed and establish trust in their ML pipelines. The structural core of enterprise MLOps relies on separating overall workflows into distinct batch training and real-time serving pipelines. Because models inevitably degrade due to "data drift," teams must enforce continuous monitoring of both technical infrastructure performance and business context accuracy. To properly manage this lifecycle, organizations utilize centralized feature stores to feed data consistently, workflow orchestrators to automate batch retraining, and model registries to version artifacts securely. When productionizing these models, decoupling the machine learning artifact from the main application via API-driven microservices or dedicated serving tools ensures greater deployment flexibility. Containerizing components with Docker and Kubernetes has become the de facto standard, though ultra-low-latency use cases may alternatively require streaming engines that seamlessly load intermediate representations like PMML or ONNX. Ultimately, successful MLOps depends less on choosing between specific vendor suites and significantly more on enforcing stringent software engineering principles across the entire data science lifecycle.

**Keywords:** enterprise mlops architecture, jupyter notebook limitations, machine learning production, continuous model retraining, data drift monitoring, cross-functional data teams, real-time model serving, batch processing pipelines, centralized feature stores, model artifact registries, api-driven microservices, kubernetes containerization, stream processing latency, pmml and onnx formats, workflow orchestration tools, automl enterprise integration

## Chapters

1. **Introduction to the speaker and data science background** (00:02) — The evolution from artificial intelligence research into modern data engineering establishes a foundational operational context.
1. **Initial challenges of machine learning in production** (01:52) — Moving beyond isolated algorithms involves establishing robust ecosystems for live operational modeling.
1. **Limitations of notebook development for production environments** (02:34) — While jupyter notebooks excel in experimentation, they lack the stability required for scalable production application.
1. **Integrating machine learning into broader enterprise software systems** (04:23) — Core machine learning code forms only a small fraction of the complex infrastructure needed for production maintenance.
1. **Adopting a DevOps culture for machine learning pipelines** (05:30) — Embracing cross-functional collaboration eliminates operational handoffs and increases trust when deploying continuous machine learning updates.
1. **Defining core roles and responsibilities in MLOps teams** (10:01) — Designing cohesive delivery groups requires integrating the distinct skills of data scientists and operational engineers.
1. **Best practices for maintaining production machine learning models** (12:45) — Treating models as versioned artifacts enables continuous monitoring and retraining against ongoing dataset drift.
1. **Architectural patterns for batch and real-time pipelines** (15:54) — Designing sustainable systems separates heavy batch training jobs from lightweight real-time prediction serving endpoints.
1. **Real-world application of MLOps architectural patterns at Wayfair** (21:24) — Combining cloud-native vendor solutions with custom internal tools successfully powers large-scale retail machine learning pipelines.
1. **Navigating the MLOps tooling landscape and vendor decisions** (22:23) — Engineering teams must carefully evaluate the tradeoffs between building custom infrastructure and adopting commercial automated solutions.
1. **Decoupling model serving from core application logic** (26:03) — Isolating models as dedicated microservices allows for independent updates without risking the entire software application.
1. **Handling near real-time model evaluation in streaming contexts** (30:09) — Embedding predictive models directly into stream processing engines strictly satisfies necessary sub-second latency performance requirements.
1. **Standardizing deployments with containers and Kubernetes orchestration** (32:43) — Packing algorithms into structured docker containers ensures reliable execution across varied cloud instances and local environments.
1. **Centralizing data preparation with a dedicated feature store** (34:07) — Deploying specialized databases unifies offline training data retrieval with high-performance real-time prediction querying.
1. **Managing batch job complexities through workflow orchestration** (36:52) — Implementing sophisticated scheduling frameworks helps teams effectively track, debug, and automate model retraining tasks.
1. **Summary and recommended educational resources for MLOps** (38:58) — Exploring dedicated organization literature and technical vendor certifications empowers practitioners to deepen architectural expertise.
1. **Audience Q&A on educational courses and freelancing prospects** (40:42) — Practical audience inquiries highlight valuable learning materials, field entry strategies, and independent consulting career navigation.

## Related Moments

- [Defining MLOps and its role in production systems](https://www.wearedevelopers.com/videos/825-mlops-on-kubernetes-exploring-argo-workflows) (from "MLOps on Kubernetes: Exploring Argo Workflows")
- [Transitioning machine learning models from notebooks to production](https://www.wearedevelopers.com/videos/501-model-governance-and-explainable-ai-as-tools-for-legal-compliance-and-risk-management) (from "Model Governance and Explainable AI as tools for legal compliance and risk management")
- [Introduction to DevOps for AI and MLOps](https://www.wearedevelopers.com/videos/1222-devops-for-ai-running-llms-in-production-with-kubernetes-and-kubeflow) (from "DevOps for AI: running LLMs in production with Kubernetes and KubeFlow")
- [Differences between traditional MLOps and GenAIOps](https://www.wearedevelopers.com/videos/1535-from-traction-to-production-maturing-your-genaiops-step-by-step) (from "From Traction to Production: Maturing your GenAIOps step by step")
- [Moving machine learning models from experimentation to production](https://www.wearedevelopers.com/videos/185-effective-machine-learning-managing-complexity-with-mlops) (from "Effective Machine Learning - Managing Complexity with MLOps")
- [Bridging the gap between model management and devops](https://www.wearedevelopers.com/videos/1152-ai-model-management-life-circles-ml-ops-for-generative-ai-models-from-research-to-deployment) (from "AI Model Management Life Circles: ML Ops For Generative AI Models From Research to Deployment")

## Related Articles

- [MLops – Deploying, Maintaining And Evolving Machine Learning Models in Production](https://www.wearedevelopers.com/magazine/115-mlops-deploying-maintaining-and-evolving-machine-learning-models-in-production)
- [MLOps – What’s the deal behind it?](https://www.wearedevelopers.com/magazine/125-mlops-what-s-the-deal-behind-it)
- [MLOps And AI Driven Development](https://www.wearedevelopers.com/magazine/82-mlops-and-ai-driven-development)
- [Stephan Gillich - Bringing AI Everywhere](https://www.wearedevelopers.com/magazine/489-stephan-gillich-bringing-ai-everywhere)

## Related Jobs

- [Staff, Machine Learning Engineer (L4)](https://www.wearedevelopers.com/jobs/ext/1202639-staff-machine-learning-engineer-l4) at **Twilio**
- [Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/1597388-machine-learning-engineer) at **ZEISS Group**
- [Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/588393-machine-learning-engineer) at **Twilio**
- [Machine Learning Engineer](https://www.wearedevelopers.com/jobs/ext/1355348-machine-learning-engineer) at **TWILIO**
- [Data Scientist](https://www.wearedevelopers.com/jobs/ext/1351648-data-scientist) at **Almedia**
- [Principal Software Engineer, Enterprise AI Platform](https://www.wearedevelopers.com/jobs/ext/1467292-principal-software-engineer-enterprise-ai-platform) at **GitHub**