Odyssey Platform # 12

🗞️ In this newsletter

🗓️ Events

•  Conf42 Platform Engineering 2025

🔦 Tool Spotlight

• MongoDB: The database that moves as fast as your code

🔍️ Deep Dive: Crossplane provider getting started 🧠

🎯 Stay Inspired - Case Studies

• Churn-Proof Your Models: MLOps with Databricks

👀 In Case you missed it

• Latest news & events in the platform engineering domain

📆 Upcoming Events

Get ready for Conf42 Platform Engineering 2025, an entirely online event premiering September 4, 2025, at 5 PM GMT. This annual deep-dive conference explores the full platform engineering landscape, spanning architecture, devex, automation, infrastructure, GenAI, and scalability, among others. Expect rich learning and connection through live keynotes, real-time discussions, and peer‑powered sessions—designed for platform engineers shaping the future.
👉 Register here

🔦 Tool Spotlight

Once seen mainly as a niche NoSQL option, MongoDB has surged in popularity as modern applications demand flexibility, speed, and scale that traditional relational databases often struggle to provide. Its document-oriented model allows developers to store complex, evolving data structures without rigid schemas—perfect for agile teams iterating rapidly.

MongoDB’s rise is also driven by cloud-native adoption. With MongoDB Atlas, teams get fully managed deployments, built-in horizontal scaling, multi-cloud support, and native integrations with AI/ML workflows. Features like change streams for real-time data, flexible indexing, and powerful aggregation pipelines make it a go-to choice for building responsive, data-driven platforms.

From powering microservices to handling IoT event streams and generative AI backends, MongoDB is becoming the default choice for teams who need a database that grows and adapts as fast as their applications.

Deep Dive: Crossplane provider getting started 🚀 

🎬 Act 1: The Prologue - What is Crossplane

Imagine if your engineers could manage cloud infrastructure the same way they manage application code: versioned, reviewed, and deployed alongside everything else. Crossplane is the open-source control plane that makes this possible, letting you provision and manage infrastructure using familiar Kubernetes APIs.

With Crossplane, you can also simplify how applications are deployed. Instead of giving developers a massive Helm values file or complex Kustomize overlays, you present them with a clean, minimal claim. The claim contains only the important details they need to set: the environment, the secrets to pull from, and the images to run. Everything else is pre-configured by the platform team, ensuring consistency, security, and speed.

✨ Act 2: The Enchanter’s Wand - Installing Crossplane

helm repo add crossplane-stable https://charts.crossplane.io/stable

helm install crossplane --namespace crossplane-system crossplane-stable/crossplane --create-namespace

🌱 What just happened?

• Added Crossplane’s helm repo for releases

• Installed Crossplane core components into your cluster

• Created the namespace crossplane-system

🪄 Your cluster is now ready to manage other clouds and services from one place.

🎯 Act 3: The Spellbook - Introducing Crossplane Providers

Crossplane’s true power comes from Providers. These are plugins that teach Crossplane how to manage specific cloud services like AWS, GCP, Azure, and more.

For example, to add AWS support:

kubectl crossplane install provider crossplane/provider-
aws:v0.50.0

🪄 Decoded:

• Adds the ability to create and manage AWS resources like S3, RDS, and EKS

• Installs the provider as a Kubernetes deployment so everything stays declarative

Providers exist for almost every major cloud, database, configuration management and SaaS service . You can run multiple providers side-by-side in the same cluster.

🔑 Act 4: The Keyring - Setting Up Provider Credentials

To connect a provider to your cloud account, you store credentials securely as a Kubernetes secret:

kubectl create secret generic aws-creds -n crossplane-system \
  --from-literal=key=YOUR_AWS_ACCESS_KEY_ID \
  --from-literal=secret=YOUR_AWS_SECRET_ACCESS_KEY

Then, create a ProviderConfig to tell Crossplane where to find them:

apiVersion: aws.crossplane.io/v1beta1
kind: ProviderConfig
metadata:
  name: default
spec:
  credentials:
    source: Secret
    secretRef:
      namespace: crossplane-system
      name: aws-creds
      key: key

This links your AWS provider to the secret and follows Kubernetes best practices for secure credential management.

🪄 Act 5: The Cloak - Claim-Based Simplicity

Here’s where the developer experience shines. Instead of having engineers maintain dozens of Helm chart variables or Kustomize patches, you give them a simple claim like this:

apiVersion: apps.example.org/v1alpha1
kind: MyAppClaim
metadata:
  name: payment-service
spec:
  environment: staging
  image: registry.example.com/payments:v1.4.2
  secretsRef: payment-service-secrets

🚀Act 6: The Business Benefits

• For Platform Teams: One API to manage all infrastructure, enforce policies, and evolve the platform without breaking developer workflows

• For Developers: Minimal configuration, faster onboarding, and fewer mistakes

• For the Business:

  • Faster time to market with self-service provisioning

  • Lower operational costs through standardisation

  • Better security and compliance with centralised governance

  • Consistency across environments to reduce downtime

  • Scalability without adding headcount

🎇 Curtain Call - Why Adopt Platforms Like This

Crossplane bridges the gap between platform engineering and developer productivity. It lets platform teams set strong guardrails while giving developers the autonomy to move fast. The result is a platform that drives innovation, controls costs, and scales with the business.

In today’s competitive market, that combination is not just nice to have, it is essential.

🎯 Stay Inspired - Case Studies

Churn-Proof Your Models: MLOps with Databricks

Who: A large enterprise deploying predictive analytics to understand and reduce customer churn across millions of users.

What They Did:
Built an end-to-end MLOps pipeline on Azure Databricks to take a churn prediction model from concept to continuous production delivery. The solution included:

• Data ingestion and preparation using Delta Lake for reliable, versioned datasets

• Feature engineering workflows that automatically refresh as new data lands

• Experiment tracking and model management with MLflow to ensure reproducibility

• Automated CI/CD pipelines via Azure DevOps for model deployment into production environments

• Model drift detection and retraining to keep predictions accurate over time

• Governance and auditability embedded into every stage to meet compliance needs

Tech Stack & Tools Used: Azure Databricks, Delta Lake, MLflow, Azure DevOps, Azure Machine Learning

Why It Matters:

✅ Operationalised AI – Turning a data science project into a reliable business capability with automated delivery
✅ Continuous accuracy – Drift detection and retraining prevent performance degradation over time
✅ Enterprise-grade governance – Ensures ML outputs meet compliance, reproducibility, and audit standards

👀 In Case You Missed It…

The AI quality bottleneck every platform team will face AI‑generated code boosts output—but it also means more bugs, broken builds, overloaded code reviews, and security risks. Without strong CI/CD pipelines, golden paths, and automated quality gates, platform teams risk collapsing under the surge of AI‑accelerated development.

Google Cloud’s New AI Agents Supercharge Data Science and Engineering
Google Cloud has introduced six specialized AI agents designed to streamline everything from feature engineering and pipeline automation to exploratory analysis and DevOps workflows, bringing an AI-native edge to enterprise data and software projects.

From YAML to Intelligence: Platform Democracy Rising
CNCF’s latest take highlights how platform engineering is shifting from centralized control toward a collaborative “Platform Democracy”; stakeholders like developers, SREs, and compliance teams are all contributing to and consuming platform capabilities, reshaping internal developer platforms (IDPs) into shared products.

Till next time,

Odyssey Platform