Complex software needs more than code.
We design and build software where architecture, business logic, integrations and reliability matter.
From data-intensive applications and SaaS platforms to real-time systems and production AI, PikiSoft brings the engineering experience to make the right technical decisions, and the hands-on expertise to implement them.
Where we usually start
Most projects start with software that holds the business back, or a technical decision too important to guess.
The system runs the business, but every change is risky and slow. We make it changeable step by step, without stopping day-to-day operations.
Legacy that blocks growth
Business logic ends up in scripts and workflows nobody owns. We decide what belongs in your product and what is operational automation, where a tool like n8n is the right fit, and build both properly.
Automation without clear boundaries
The demo works; production is something else. We build AI in with evaluations, human oversight, cost tracking and monitoring.
AI that has to run reliably
A platform choice, multi-tenancy, build versus buy: decisions that last for years. We weigh the options and risks, and build what we recommend.
Architecture decisions with consequences
What we do
Custom Software & Product Engineering
Software built around complex business and technical requirements, designed to scale and grow with the business.
- Requirements, data models and an architecture that grows with users, data and teams
- Interactive applications with demanding calculation and business logic
- SaaS platforms and backend systems, from APIs to infrastructure
- TypeScript, React and Next.js; Node.js or Python on PostgreSQL
- CI/CD, preview environments and maintenance after launch
- Product engineering
- Data models & APIs
- React & Next.js
- Node.js & Python
- PostgreSQL
- Cloud & CI/CD
Systems Architecture & Integration
Designing architectures and making existing systems work together.
- Architecture for multi-tenant platforms and distributed services
- API integrations between CRM, ERP, accounting and in-house systems, with retries and alerts
- Webhooks and event-driven processing instead of nightly batches
- Operational automation with clear boundaries: what belongs in the product is built into it, the rest runs on a workflow platform such as n8n, self-hosted, version-controlled and monitored
- Integrations with external platforms and partners
- Architecture
- Multi-tenancy
- API integration
- Event-driven
- Workflow orchestration
- Partner integrations
AI Engineering
AI in real applications and business processes: with evals, guardrails and data that stays in the EU.
- Knowledge management: structured knowledge that people and AI agents share, with retrieval (RAG) over your own documents and sources users can check
- LLM features inside your existing application and data
- Agent orchestration through defined tools and MCP servers
- Guardrails and human approval for actions that matter
- Evals, cost tracking, logging and monitoring before and after release
- Set up with GDPR and the EU AI Act in mind: EU data residency, retention periods and a register of AI vendors
- Knowledge management & RAG
- Tool calling
- Agent orchestration
- Model Context Protocol
- Guardrails
- Human-in-the-loop
- Evals & observability
- GDPR & AI Act
Technical Leadership & Delivery
Technical leadership with the delivery capacity to match: from discovery to deployment, scaled to what the project needs.
- Evaluating architectural options and identifying risks
- Technical direction and roadmap
- Interim CTO or technical lead roles
- Build-versus-buy decisions
- Assessing AI readiness: which processes suit AI, what data it needs, what it costs and what the risks are
- Staying involved in the engineering, from discovery to deployment
- Architecture reviews
- Interim CTO
- Risk assessment
- AI readiness
- Build vs buy
- Delivery planning
Selected work
Projects where we owned architecture and delivery.
Full ownership of the software lifecycle: from product decisions and architecture to CI/CD, security and operations.
- Product prioritisation and a roadmap, with every significant choice recorded as an architecture decision record
- Core architecture: one domain layer behind both a REST API and an MCP server, plus the design and build of the mobile-first web client
- Quality built in: unit, end-to-end and mutation tests, plus evals for agent behaviour
- CI/CD with staging and production environments
- Security reviews, passkey sign-in, data export and account deletion, monitoring, alerts and backups
Build a multitenant voice AI platform that telecom partners can offer inside their own platforms.
- Built the multitenant reseller portal and moved the platform to a SIP-native architecture, so a telecom partner could offer the AI agents inside its own platform
- Designed the capacity model behind guaranteed concurrency for larger enterprise customers
- Ran the enterprise agents after go-live: a per-customer success dashboard and Promptfoo regression evals starting from 150+ real calls
- Replaced manual releases with CI/CD and observability (logs, alerts, Sentry), and set up an AI vendor register for data residency and retention
Deliver data-heavy internal products across several teams, with frontend and backend working in parallel.
- Owned frontend architecture for forecasting, quotation and inventory applications
- Worked with product, design and backend teams on architecture, estimates and plans
- Introduced preview environments so stakeholders could review changes before release
- Introduced API mocking so frontend and backend could ship in parallel
Why PikiSoft
Experience beyond individual features.
I've been fascinated by software since I was a kid, when I discovered that even a calculator could be programmed, and later started experimenting with Visual Basic macros in Excel. That curiosity led me to a degree in computer science, and it hasn't faded: today I'm just as excited about what AI makes possible.
Over the past 15+ years I've worked across countries, teams and industries. I extended engineering teams and built products, did platform engineering at an established Dutch fintech company, built a multi-market rendering strategy for a global sports streaming platform, and built multilingual voice AI agents that hold natural conversations with people in their own language.
I started PikiSoft in 2023 because I wanted to work closer to the actual problems, with more freedom to make decisions and take responsibility for the results. I bring that same approach to client projects and my own products: understanding the problem, challenging assumptions, making thoughtful engineering decisions and staying involved from the first idea to working software.
- Fintech
- Digital banking platform
- Streaming
- Global sports streaming
- Enterprise
- Forecasting, quotation and inventory
- Voice AI
- Multilingual agents in production
- 01
Start with the problem
We map the requirements, existing systems, technical constraints and expected outcome before selecting a solution.
- 02
Make informed decisions
We evaluate architectural options, dependencies and trade-offs. What happens when systems grow, integrations fail or requirements change?
- 03
Build and verify
We turn architectural decisions into working software, validate assumptions, test integrations and address problems that emerge during implementation.
- 04
Think beyond delivery
We consider how the software will be deployed, operated, maintained and extended.
What to expect
Generative AI, built for production.
Generative AI is changing how software is built, delivered and used. It opens up new opportunities for automation and changes what businesses can achieve with technology.
But making AI useful in real-world operations takes more than integrating a model. Systems need to be secure, reliable and transparent, with clear accountability and control over decisions and outcomes.
The key is knowing where AI's flexibility adds real value and where predictable, rule-based software should remain in control. Getting that balance right is what makes AI sustainable in production.
AI and data sovereignty
The standards behind every project that handles data or AI, and the principles our own product is built on.
EU hosting and sovereignty
Applications and data run in European data centres. Wherever possible, we choose European-owned cloud providers, so your data stays under European law.
Your data, your accounts
Code, data and cloud environments are in your name, not ours. A data processing agreement and registers of your data and AI vendors show what is kept where, for how long, and who processes it.
Autonomy under control
Together we decide how much autonomy the AI gets for each task, based on the risk. Guardrails and evals keep it within those limits, and when it is unsure or stuck, a person takes over.
waybr8
An AI-native workspace for shared operational knowledge.
PRIVATE BETA
While building agentic systems, we kept coming back to a fundamental question: how can people and AI agents share knowledge, build on each other's work, and keep everything understandable, trustworthy and under control?
That question became waybr8: a shared workspace where people and AI agents work with the same structured knowledge, with clear permissions, traceable changes and human oversight where it matters.
What started as a solution to our own engineering challenges has grown into a product that now powers our day-to-day operations.
Private beta: access on invitation.
Clear, actionable knowledge
Structured pages, tasks and decisions: easy for people to understand, unambiguous for AI agents, and ready to act on.
MCP integration
Connect the agents you already use, such as ChatGPT, Claude Code, Grok or ElevenLabs, through the Model Context Protocol.
Access, history and approval
The same roles and permissions apply to people and agents. You see which agent changed which part of your knowledge, and once a person approves a note, that version is locked.