Skip to content

Development

Custom software development, agent-assisted

Custom software and MVPs, built with AI agents. Human architecture, machine velocity.

Software projects often lose months to hand-offs between specialists and to guesswork before the first release. Interlusion builds software differently: autonomous agents handle code generation, testing, refactoring, and deployment under human supervision. The architect designs the system. The agents execute. The result: months of guesswork become weeks of measurable progress.

This isn't a theoretical approach. It's how Bettles — a full-stack application with UI, backend, and infrastructure — was rewritten from scratch, solo, in about six months. The same methodology applies to every client engagement.

Every project starts with architecture, not code. System boundaries, data flows, and deployment strategy are defined before a single line is written. Agents then execute within those guardrails — fast, consistent, and testable. Humans review every merge, make judgment calls, and own the design decisions.

The outcome for clients: fewer hand-offs, shorter feedback loops, and one partner from architecture to launch.

Capabilities

What Interlusion builds

Seven areas of deep technical capability — each backed by 25+ years of engineering experience and refined through agent-assisted workflows.

Software Architecture

Distributed systems with clean architecture, built for scalability and long-term maintainability. The architecture comes first; the agents work within it.

Agent-Assisted Software Development

Full-stack applications built by AI agents under human supervision — from UI to backend to infrastructure. Clean architecture, tested code, production-ready from day one. Agents handle the repetitive work. Engineers focus on design decisions.

Rapid Product Prototyping

From idea to working MVP in weeks, not months. Orchestrated agent workflows eliminate guesswork and collapse traditional development timelines. Validated concepts with real code, not slide decks.

Cloud & DevOps

Infrastructure as code, CI/CD pipelines, containerisation, scalability architecture. GitOps-driven continuous delivery that runs itself once configured. Kubernetes, Terraform, and automated monitoring from the start.

Web3 & Decentralised Systems

Token-based user experiences and blockchain-enabled services. Practical decentralisation where it adds real value — not hype, architecture.

Industry 4.0 / IoT

Connected devices, real-time dashboards, sensor integration. Industrial systems that surface actionable data and reduce operational friction across the factory floor.

Enterprise Applications

Complex browser-based business applications, typically built with Angular. Designed for scale, maintainability, and long-term evolution. Security and compliance baked in from day one.

Process

How a project unfolds

Every engagement follows the same four phases, with clear milestones.

01

Discovery & Architecture

Deep dive into requirements, constraints, and business context. System boundaries are defined, data flows mapped, and a technical architecture is drafted. This phase prevents expensive pivots later.

02

Foundation Sprint

Core infrastructure, CI/CD pipelines, authentication, and the base application scaffold go live. Agents generate the boilerplate. Humans review and refine the architecture. A deployable skeleton exists within days.

03

Iterative Build

Feature development in short cycles. Agents write, test, and refactor code under supervision. Each cycle produces a working increment — visible progress, not status reports. Weekly demos keep stakeholders aligned.

04

Launch & Handover

Production deployment, documentation, knowledge transfer. The system runs on its own infrastructure with monitoring and alerting. Interlusion provides post-launch support during the stabilisation period.

Engagement Models

Work the way that fits

Different projects need different structures. Interlusion adapts to what the situation requires.

4–8 weeks

MVP Sprint

Concept to working product. Ideal for startups validating an idea or established companies testing a new direction. Includes architecture, core features, deployment, and a launch-ready application.

Best for: Startups, innovation teams

2–6 months

Project Partnership

End-to-end product development with defined scope and milestones. Weekly demos, continuous deployment, and full documentation. Interlusion embeds as an extension of the client team.

Best for: CTOs, engineering leads

Monthly

Ongoing Retainer

Continuous development, maintenance, and feature work. Predictable capacity, flexible priorities. Best for products that need sustained engineering attention after initial launch.

Best for: Established products

Technology

Tech stack

Stack-agnostic by principle, opinionated by experience. With agents writing most of the code, the client's ecosystem decides: if a team runs Java/Spring, .NET or anything else, Interlusion works in it. What Interlusion brings is architecture, engineering discipline and the agent workflow, not a framework.

Where Interlusion chooses, it uses a small set of proven defaults that agents produce reliably good results with: strong typing, fast feedback, predictable conventions.

Frontend
TypeScript, Angular for complex applications, Astro for sites, Tailwind CSS
Backend
Go, Kotlin, Node.js/NestJS, Python for AI workloads, PostgreSQL
Infrastructure
Kubernetes, Terraform, GitOps (Flux CD), Hetzner Cloud, Cloudflare
AI
Model-agnostic: commercial APIs (Anthropic, OpenAI) and open-weight models, in the cloud or on-premises

Proof Point

Bettles — Rewritten Solo in About Six Months

A complete full-stack application — UI, backend, infrastructure — designed and built by a single architect using agent-assisted workflows. From the first commit in December 2025 to live in June 2026. The same methodology is available to every Interlusion client.

  • UI, backend and infrastructure
  • Agent-assisted development throughout
  • Solo architect, production-grade result
  • Live since June 2026

Common questions

AI agents handle code generation, testing, refactoring, and documentation under human supervision. The architect designs the system, defines constraints, and reviews every change. Agents execute the implementation within those guardrails. Think of it as a senior engineer directing a team of tireless, consistent junior developers.

A typical MVP sprint runs 4 to 8 weeks from kickoff to deployment. The exact timeline depends on scope and complexity. Agent-assisted workflows compress the build phase significantly, but discovery and architecture still take the time they need — cutting corners there creates expensive problems later.

Every project is built with scale in mind from the start — containerised deployments, infrastructure as code, and clean architecture patterns that support growth. Interlusion offers ongoing retainer engagements for continued development, or can hand over a well-documented codebase to an internal team.

Absolutely. The project partnership model is designed for exactly this. Interlusion integrates with existing workflows, repositories, and communication tools. The goal is to accelerate the team, not replace it.

Nine, across 25+ years of engineering: e-mobility and charging infrastructure, insurance, automotive, healthcare and care, shipbuilding, logistics, smart city and IoT, education, and sports. The approach is stack-agnostic: the client's ecosystem decides.

Have an idea?

From concept to production — Interlusion can take it the whole way. Start with a conversation.