Engineering · Stack
The stack, with reasons.
A technology list is easy; the discipline is keeping it short. Every tool below earns its place in production - here is what each one does for our clients, and why it beat the alternatives. The architecture page shows how they fit together.
Frontend
React
The default for product interfaces: component model, ecosystem depth, and hiring reality. Used in every web product we ship, including FinCalix.
Next.js
When a product needs server rendering, SEO-critical pages or hybrid static/dynamic delivery. Content-forward platforms like Terravion are the typical case.
Flutter
One codebase, both app stores, near-native performance. The honest trade for almost every product - reasoning on the mobile page.
Backend
Node.js
Fast to build, easy to staff, excellent for API services and real-time work. Shares a language with the front end, which keeps small teams quick.
Python
The language of the AI ecosystem and of data work. Every RAG pipeline and model integration we ship has Python close to the model.
FastAPI
Typed, fast, self-documenting APIs in Python. Validation at the boundary and OpenAPI docs for free - the framework our AI services are built on.
Intelligence
OpenAI
GPT models for generation, extraction and reasoning at scale. Broad capability, strong tooling. One of three providers behind our single internal interface.
Anthropic Claude
Long-context reasoning and careful instruction-following - our pick for document-heavy and review workloads, including the AI review gate in our own pipeline.
Google AI (Gemini)
Multimodal strength and competitive economics. Provider choice is a routing decision per use case, not a religion - see AI Product Development.
Cloud
AWS
The broadest service catalogue and region coverage. Default when clients have no prior footprint.
Azure
The natural home for organisations living in Microsoft identity and tooling; strong enterprise integration.
Oracle Cloud (OCI)
Aggressive pricing for compute-heavy workloads; a serious option we deploy when economics decide.
Google Cloud (GCP)
First-class data and AI services, clean Kubernetes story. Favoured when analytics and AI dominate the roadmap.
Platform
Docker
Every workload ships as a container: identical from laptop to production, portable across all four clouds.
Kubernetes
For products with many services and real scaling needs. We deploy it when justified and say so when it is not.
Terraform
All infrastructure declared as reviewable code. Environments are reproducible; configuration is documentation.
GitHub Actions
The pipeline behind every release: tests, static checks, AI review, staged rollout - shown on the architecture page.
Data
PostgreSQL
The source of truth in every product: relational rigour, row-level security for tenancy, pgvector for AI retrieval. One database engine, deeply known.
Redis
Hot paths, sessions, rate limits and lightweight queues. The difference between fast and fragile under load.
Nginx
TLS termination, routing, rate limiting and static delivery at the edge of every deployment.
Stack questions?
Ask which of these fits your product - or challenge a choice. Engineers answer, within 24 hours.
Related: Architecture gallery · Process & standards · All services · Client work