Software Engineering Daily – "Complex Workload Deployment with Will Stewart"

Date: August 21, 2025
Host: Shawn Falconer
Guest: Will Stewart (Co-founder & CEO, Northflank)

Overview

This episode dives into the contemporary challenges of deploying complex software workloads in the cloud, focusing on developer experience, infrastructure abstraction, platform engineering, and the evolution of internal tools. Shawn Falconer speaks with Will Stewart about how his startup, Northflank, aims to streamline application deployment across clouds and on-prem environments, avoiding the notorious "graduation problem" faced by Heroku-like platforms. The discussion covers technological trends, enterprise needs, abstractions over Kubernetes, developer self-service, and the future of DevOps and platform engineering.

Will Stewart’s Journey: From Game Servers to Enterprise Infrastructure

Timestamps: [00:53]–[03:41]

• Will’s early fascination with building things, starting with running and deploying game servers at university.
• The discovery that the problems they solved for game server deployment mirrored enterprise deployment needs—particularly around containers and microservices.
• Importance of accelerator programs for founders in non-tech hubs:

  "I was in a small city in the north of England called Lincoln... being able to join this accelerator and learn what building a startup was all about was great." [01:45]

• Realization, while on placement at a digital agency, that multi-customer, multi-staging environments with manual deployment steps had massive inefficiencies.

Developer Experience and the Complexity of Modern Cloud Deployments

Timestamps: [03:41]–[07:33]

• Game servers as a “gateway drug” to learning about infrastructure and Kubernetes:

  "Game servers are a gateway drug to cloud infrastructure and Kubernetes because I could spin up a Mesos cluster and deploy containers fairly simply. But it was challenging to get the control plane running." [03:58]

• Pain points: slow environment spin-up, lack of CI/CD, cumbersome manual infrastructure provisioning, inconsistent developer experience.
• Inspiration from Heroku’s developer experience, noting its weaknesses for complex or large-scale apps.

Why Deployment Got (Even) More Complicated

Timestamps: [07:33]–[09:57]

• The proliferation of tools and the challenge of integrating multiple open source solutions:

  "It's very easy to start a kubernetes cluster, it's very easy to run a helm install, but it's very hard to then integrate all of these very easy things into one consistent unified platform." [08:33]

• The persistent internal platform problem: each company with decent engineering resources ends up stitching together similar infrastructure, resulting in inefficiency and redundant effort.

The Heroku "Graduation Problem" and Building a Flexible Platform

Timestamps: [09:57]–[12:54]

• Heroku made things easy, but scaling and specific needs trigger the need to "graduate" to more configurable—but more painful—solutions.
• Northflank’s abstraction: providing a self-service developer platform that bridges the gap between Heroku (easy) and bring-your-own-infrastructure (flexible but hard):

  "Northlank is trying to solve this problem in the middle... teams need a self-service developer platform that if you squint Cloud Foundry and Heroku look fairly similar." [11:19]

• Using Kubernetes as the "operating system" foundation to provide portable, cloud-agnostic deployment.

What Using Northflank Looks Like

Timestamps: [12:54]–[16:17]

• Setup:
  • Sign up, connect cloud account, control plane and runtime separation.
  • Can spin up a Kubernetes cluster and deploy workloads in under 30 minutes, without needing to write Kubernetes YAML or know its internals.

  "Within 30 minutes you could have an EKS cluster created inside your AWS account and your workloads deployed to production without ever having to know what a Kubernetes cluster is or write a line of YAML." [13:57]

• CI/CD:
  • Connect version control (GitHub, Bitbucket, etc.), automatic deployment flows.
  • Per-pull-request preview environments, auto spin-down for cost-saving.
• Local Development:
  • Not local-first, but allows proxying remote containers into a local dev setup.

  "We actually call Northlake a post commit platform. So when you've pushed a version control, Northlake handles everything..." [15:52]

• Deployment Models:
  • SaaS (control plane and runtime managed by Northflank)
  • Hybrid (control plane in Northflank, runtime in user cloud)
  • Fully self-hosted (both control/runtimes in customer infra, in development for air-gapped/enterprise needs)

The Future: Bring Your Own Cloud, Control Plane Separation, Enterprise Demands

Timestamps: [17:49]–[19:27]

• Universal trend: customers, especially enterprises, demand to run vendor software in their own cloud or on-prem, for security/compliance/resource utilization.

  "Enterprise can't buy your software if they can't deploy it themselves. And if you haven't separated control plane and runtime, you better start thinking about it." [18:25]

• Northflank’s productization for SaaS builders: enabling their customers to offer “bring your own cloud” without building deep multi-cloud infra themselves.

Technical Deep Dive: Control Plane, Runtime, DSL Primitives

Timestamps: [20:11]–[25:45]

• Control Plane:
  • Holds workload configurations, listens to changes, applies/updates manifests, auto-heals, maintains drift state, supports GitOps and UI-driven changes.

  "Our control plane is listening and subscribing to all of the Kubernetes clusters. It's observing when things start failing, tries to auto heal..." [20:52]

• Runtime:
  • Executes containers, ensures network/service mesh health; does not require custom agents, all via Kubernetes API for portability.
• Infrastructure as Data, not Code:
  • Simpler abstractions—shift away from Terraform/HCL to declarative JSON.
  • Reduces 1,000s of lines of traditional config to ~130 lines to define a multi-service, redundant stack.
• Custom Extensibility:
  • Support for “bring your own add-on” (Helm charts, etc.) for new data services.

  "Our job is to provide the right abstractions at a workload level and then we will be able to change in and out the ultimate underlying Orchestrator when it comes along." [26:27]

Databases and Stateful Workload Management

Timestamps: [24:03]–[25:45]

• Six built-in managed stateful offerings: Postgres, Redis, MongoDB, RabbitMQ, MySQL, Minari.
• For unsupported DBs or cloud-managed DBs, simply inject connection secrets.
• Operator pattern for stateful workloads inside Kubernetes.
• Flexible: happy if users want to bring their own data service or use a cloud-managed DB.

Abstraction, Flexibility, and the Future of Kubernetes

Timestamps: [26:08]–[27:09]

• Northflank abstracts Kubernetes primitives and aims to enable flexibility across future changes in orchestration technologies.
• Provides easy extensibility (via “add-ons”) for anything not natively abstracted.

From Game Servers to Serious Enterprise: Reliability and Team Practices

Timestamps: [29:00]–[31:43]

• The criticality of enterprise software: downtime means real financial loss for customers.

  "If our software goes down, our customers lose money. And that's something we try and ingrain in all of our engineers..." [29:39]

• Northflank’s conservative approach to new features; never deprecated a feature in six years, builds with an eye on long-term reliability and integration.

Adapting to New Tech: When to Add Features (GPUs, WASM, etc.)

Timestamps: [31:59]–[33:41]

• Careful adoption of new deployment primitives (e.g., GPUs) only when there's real customer demand. Skipped WASM/serverless preemptively.
• GPU support added as just another primitive, enabled Northflank customers to quickly deploy GPU or CPU workloads side-by-side.
• Case study: Two engineers scaling to millions of users using Northflank, compared to much larger teams barely managing lower loads elsewhere.

The Evolving Role of DevOps and Platform Engineering

Timestamps: [34:25]–[35:55]

• The trajectory: sysadmin → DevOps → platform engineer.
• Companies can’t justify large teams solely dedicated to building internal platforms; these efforts are repeated, take years, require frequent re-platforming.

  "DevOps is going to be seen as a cost center, not a relief... If you're not driving the critical business logic of the business, there'll be questions there." [35:20]

• Emergence of platforms like Northflank enables businesses to focus resources on actual product development.

The Future of Cloud Infrastructure Management

Timestamps: [35:59]–[37:04]

• Cloud (public and private) is increasingly a commodity; differentiation happens at the abstraction/interface layer, not the underlying provider.
• Trend towards hardware-agnostic, developer-friendly interfaces:

  "The cloud provider becomes less relevant because ultimately... it's what you run on top of that that matters." [36:39]

What’s Next for Northflank

Timestamps: [37:04]–[39:04]

• Building a self-deployable control plane for enterprise, aiding procurement and air-gapped use cases.
• Advanced GPU features: on-demand, self-service spot GPUs.
• Ongoing push to replace legacy platforms (Cloud Foundry, OpenShift, etc.) with a modern, out-of-the-box alternative for both startups and enterprises.
• The importance of supporting both startup-like self-serve and heavily regulated, air-gapped enterprise environments.

Memorable Quotes

• "Customers don't pay you to write YAML."
  — Will Stewart, [08:53]

• “If our software goes down, our customers lose money. And that’s something we try and ingrain in all of our engineers...”
  — Will Stewart, [29:39]

• “There are a lot of cloud infrastructure [tools], there are so many tools now in the last 10 years but ironically it's if not more complex than it was 10 years ago to ship a container.”
  — Will Stewart, [06:48]

• “DevOps is going to be seen as a cost center, not a relief. And teams are going to have to adapt to that.”
  — Will Stewart, [35:20]

Notable Moments and Case Studies

• [33:41] Will shares a story about a company (Weights) where two engineers are managing millions of users and large-scale CPU/GPU deployments using Northflank—a testament to the power of good abstractions and automation.

Closing Thoughts

Will Stewart provides an insightful, candid view into the evolution of cloud workload deployment, touching on both the technical and organizational challenges. Northflank positions itself as an enabling abstraction over the flaky, piecemeal world of open-source infra and custom internal platforms, focusing instead on developer experience, cloud agnosticism, and scalable operations. The episode is valuable for anyone interested in DevOps, platform engineering, cloud infrastructure, or SaaS business dynamics.