Snyk CTO: Platform engineering is a DevOps ally in the AI era
This is a guest post for the Computer Weekly Developer Network written by Danny Allan in his role as chief technology officer at Snyk.
Allan writes in full as follows…
There’s a common misconception that platform engineering is here to replace DevOps, but I believe it offers the potential to empower it. A well-crafted set of practices and tools amplifies what DevOps teams can achieve – and the collaborative culture that DevOps fostered remains as essential as ever.
What’s changing is how we deliver on that culture.
Platform engineering builds on the principles of DevOps, addressing challenges by providing standardised, scalable platforms with integrated tools and services. In other words, platform teams create the paved roads that make DevOps practices easier to adopt across an organisation.
This alliance is especially critical in today’s AI-driven development landscape. Modern software delivery isn’t just pushing app code; it increasingly involves machine learning models, data pipelines and cloud AI services. DevOps needs a boost to handle this complexity and platform engineering provides exactly that.
It’s telling that some 94% of organisations now see AI as “critical” or “important” to the future of platform engineering, while 86% say platform engineering is essential to realising AI’s full business value. At Snyk, we’ve seen AI initiatives accelerate when a solid platform is in place. Rather than sidelining DevOps, platform engineering is the partner that equips teams to ride the AI wave.
Self-service (with best practices)
Traditionally, engineers had to wrangle disparate tools for CI/CD, configuration management, monitoring and more. Platform engineering instead offers a central, pre-integrated toolbox, reducing cognitive load by giving DevOps teams a central hub of tooling and workflows chosen with their needs in mind.
Crucially, these self-service platforms also embed automated security checks and reliability measures by default. Efficient security can never be an afterthought – it must be part of the pipeline and an internal platform makes integration invisible yet effective.
Automated vulnerability scanning, automated dependency updates and supply-chain analysis are key to making security a consistent and integral part of the development process. In other words, the platform acts as a safety net, catching vulnerable packages or misconfigurations early to introduce security at inception. This embedded DevSecOps means developers can move fast without breaking things as the platform enforces security policies and compliance in the background.
Platform engineering also delivers consistency. By standardising environments and workflows, it eliminates the “works on my machine” syndrome and drift between teams. If every microservice team deploys via the same container baseline and CI/CD template, for example, you get uniformity in how apps are built and run.
Enabling ML workflows for DevOps
Modern AI/ML development introduces specialised needs that classical pipelines weren’t built for.
AI-ready Kubernetes clusters are one example. Training and serving machine learning models often requires GPU acceleration, high-memory nodes, or distributed processing frameworks. Instead of each data scientist or ML engineer filing tickets and waiting weeks, the platform team can offer a pre-configured “ML sandbox” cluster optimised for AI workloads.
Another key component is an ML model registry. DevOps is accustomed to artifact repositories for build artifacts or container images. In the AI world, the “artifacts” are trained models (and their lineage). A model registry is a version-controlled hub where models are stored, validated and tagged. Leading teams treat this as a first-class feature of their platform. The platform can also simplify automated data pipelines which are vital for feeding AI models. Data ingestion, validation routines and feature-engineering workflows can be provided as on-demand services. This integration means DevOps teams don’t have to manually wire separate ML infrastructure – the platform extends their CI/CD to cover AI workflows too.
In essence, platform engineering makes AI/ML just another user of the DevOps pipeline, rather than an outlier. We provide the mission control centre for AI initiatives: high-performance infrastructure (GPUs, big data storage), data management and model-serving environments all packaged in a manageable, self-service way.
This approach enables data scientists and ML engineers to develop, train and deploy models efficiently – and it enables DevOps to absorb AI into its domain without being overwhelmed. When an internal platform supports AI/ML out of the box, launching a new machine learning feature is as straightforward as deploying a microservice.
Acknowledging the challenges
Snyk’s Allan: A platform promising “golden paths” can become a constraint if it’s too prescriptive.
Of course, no technology or methodology is a silver bullet and platform engineering is no exception. There are several recurring concerns.
For starters, building an IDP requires dedicated engineers to build and maintain the tooling, automation and infrastructure.
The key is incremental progress: start by solving one high-pain point, demonstrate value and only then expand.
A platform promising “golden paths” can become a constraint if it’s too prescriptive. Developers fear being stuck in a “cattle chute” – a process that stifles creativity. Offer extensibility so teams can plug in custom steps or choose from approved tech stacks and evolve the platform based on feedback.
Misalignment with developer needs is also a worry. Developers will bypass a platform that doesn’t solve their pain points. You have to treat developers like customers and earn their buy-in. This means evangelising benefits, onboarding gradually and maintaining an open feedback loop to learn where friction exists.
Human-centred AI-enhanced future
Far from diminishing the role of developers, the trio of AI, platform engineering and DevOps can elevate human creativity and productivity. By offloading repetitive work and providing smart automation, platforms free developers from being YAML experts or cloud sages for every task. We still need developers to decide why a feature exists, how it behaves and when it’s ready to ship. The platform clears the runway; the developer flies the plane.
Rather than viewing AI as a threat, I see it as a liberator. If a platform uses AI to auto-tune CI/CD pipelines or analyse logs, engineers spend less time firefighting and more time building value. It’s a virtuous cycle: AI and automation enhance the platform; the platform enhances DevOps; DevOps enhances developer innovation.
The result: faster, higher-quality software delivered by empowered, resilient teams. What’s not to like?
