Backend and API engineering
Modern PHP and TypeScript service layers, versioned REST APIs, authorisation policies, queued work and the resource contracts that keep clients from breaking on every release.
digitalcosmos.io is an independent engineering practice based in Haarlem, the Netherlands, working with clients across Europe.
digitalcosmos.io is led by Dariusz Dusza, a lead software engineer whose career has covered the full length of the stack rather than a slice of it: Linux systems and networks first, then full-stack product development, then architecture and technical leadership on distributed cloud platforms. The order matters. Starting underneath the application and working upwards leaves you with an instinct most engineers never develop: the ability to look at one slow request and know whether the answer lives in the query planner, in the container, or in the meeting where somebody agreed to the requirement.
That range is still being extended rather than coasted on. The same engineer who once tuned kernel and database parameters now runs self-hosted language models, vector search and retrieval pipelines in production, because the only reliable way to learn what a technology truly costs is to be the one carrying its pager at three in the morning. Nothing is recommended here that has not been operated here.
We stay small on purpose. Every engagement is handled by the person who designed it, which keeps the feedback loop short and makes it impossible to hide a bad decision behind a delivery process. Client work spans early-stage products that need a first architecture, established platforms that need to be modernised without being switched off, and organisations that simply need someone senior to be responsible for the systems they depend on.
Alongside client work we design, build and operate our own multi-tenant SaaS platform end to end: application, mobile builds, four production databases, self-hosted AI services, cloud infrastructure, billing, email and compliance. Running our own product at production scale is what keeps the advice we give grounded in consequences we personally live with.
Founder and lead engineer
Works across application code, distributed systems, cloud infrastructure and applied AI, with a standing bias towards systems that stay boring under load. Still writes the difficult parts personally.
Lead software engineer at Flexpedia
Technical leadership on platform architecture, delivery process and infrastructure reliability.
Previously senior software engineer at Harlem Next
Full-stack product development and the operational side of shipping it continuously.
Earlier: PHP and JavaScript developer, and Linux system administrator, at Loginet BV
Where the infrastructure half of the job started: servers, networks, backups and the pager.
Engineer's degree in information technology, Wyższa Szkoła Informatyki i Zarządzania w Bielsku-Białej
A formal grounding in the fundamentals that outlast frameworks.
Based in Haarlem, the Netherlands
Working remotely across Europe, on site in the Randstad when it helps.
Not a list of everything ever touched, but the areas where the experience is deep enough to be worth paying for.
Modern PHP and TypeScript service layers, versioned REST APIs, authorisation policies, queued work and the resource contracts that keep clients from breaking on every release.
Component-driven single-page applications in TypeScript, with shared design systems, strict typing, accessibility and multilingual delivery treated as build-time guarantees.
Infrastructure as code, container platforms, self-managed database and cache servers, replication, snapshots, blue-green releases and the maintenance windows nobody notices.
Tool-using agents, retrieval pipelines with hybrid search and rank fusion, embeddings and vector stores, evaluation harnesses scored with real information-retrieval metrics, self-hosted model servers, and machine-learning inference running in production.
Multi-database architectures, database-enforced tenant isolation, audit stores separated from operational data, and migration strategies safe to run on a live system.
Agile delivery without cargo cult, code review culture, mentoring, and being the person who has to make the call when two good options disagree.
Clean architecture, readable code and reviewable infrastructure are not aspirations here, they are the delivery standard. Every project ships with the pipelines, tests and documentation needed to change it safely a year from now.
Experience spans interface code and kernel-adjacent operations, so systems are designed for how they will actually behave under load, failure and cost pressure rather than only for the happy path.
Having moved from individual contributor to technical lead, the perspective covers both what a team needs to move fast and what a client needs to trust the result. Both require the same thing: honesty about status.
Tenant isolation, audit trails, least privilege and GDPR-grade data handling are part of the initial design. Retrofitting any of them costs several times what including them does.
digitalcosmos.io exists at the point where solid software engineering meets infrastructure that has to survive contact with real users. You get one accountable partner who designs the system, writes the important parts, deploys it, and stays reachable when it matters.
What that means concretely: