Boring where it counts
Novel technology is reserved for the part of your product that is genuinely novel. Everything else runs on tools with a decade of operational history.
Technologies
Multi-stack on purpose
We do not have a house framework to sell you. The right technology is the one that fits your problem, your team's skills and your hiring market once we have handed over.
How we choose technology
Hire-able by you
We will not leave you with a stack you cannot recruit for. If your market has ten React developers and one Elm developer, that decides it.
Exit paths documented
Every significant dependency comes with a written answer to "what would it take to replace this?" — before it is adopted.
The stack
What we build with
Grouped by practice. Anything listed here is running in production for a client today — we do not list technologies we have only piloted.
Interfaces built to a performance and accessibility budget, on frameworks chosen for the product rather than for fashion.
React
Primary UI library
Next.js
SSR & edge rendering
TypeScript
Type safety end to end
Vue
Nuxt & composition API
Svelte
Lightweight interfaces
Tailwind
Token-driven styling
Vite
Build tooling
Storybook
Component documentation
Services designed for throughput, observability and a data model that survives years of change.
Node.js
NestJS & Fastify
Python
FastAPI & Django
Go
High-throughput services
PHP
Laravel & modern PHP 8
Java
Spring Boot
.NET
Enterprise integrations
GraphQL
Federated schemas
gRPC
Service-to-service
Cross-platform where it fits, native where the product demands it — benchmarked on real low-end devices.
Flutter
Default cross-platform
React Native
Shared web expertise
Swift
Native iOS
Kotlin
Native Android
Firebase
Messaging & analytics
Fastlane
Release automation
Retrieval, forecasting and analytics built with evaluation harnesses so quality is a number rather than an opinion.
Claude API
Production LLM workloads
PyTorch
Model training
pgvector
Embedded vector search
Pinecone
Managed vector store
ClickHouse
Analytical workloads
dbt
Warehouse modelling
Airflow
Orchestration
MLflow
Experiment tracking
Reproducible infrastructure, progressive delivery and cost governance built in before the first invoice surprises anyone.
AWS
Primary cloud platform
Azure
Enterprise & compliance
Google Cloud
Data & ML workloads
Kubernetes
Container orchestration
Terraform
Infrastructure as code
ArgoCD
GitOps delivery
GitHub Actions
CI/CD pipelines
Cloudflare
Edge & security
The right store for the access pattern, with migration paths planned rather than discovered.
PostgreSQL
Default relational store
MySQL
Legacy & LAMP estates
Redis
Cache & queues
MongoDB
Document workloads
DynamoDB
Serverless scale
Elasticsearch
Search & logs
Kafka
Event streaming
S3
Object storage
Gates that fail the build, and telemetry that pages a human only when a human is needed.
Playwright
End-to-end testing
Vitest
Unit & integration
k6
Load testing
axe
Accessibility gates
Datadog
Metrics & tracing
Sentry
Error monitoring
Grafana
Dashboards & SLOs
Snyk
Dependency scanning
Research and systems tooling that produces components engineers can build from without guesswork.
Figma
Design & prototyping
FigJam
Workshops & mapping
Style Dictionary
Token pipelines
Maze
Usability testing
Framer Motion
Interface motion
Lottie
Vector animation
Engineering standards
Gates that fail the build
Standards written down but not enforced are preferences. Ours run in the pipeline, and a pull request that breaches them does not merge.
- Performance budget. Largest Contentful Paint, bundle size and interaction latency thresholds per project.
- Accessibility gate. Automated axe checks plus a manual assistive-technology pass before each release.
- Security scanning. Dependency, container and secret scanning on every commit, with signed build artefacts.
- Test coverage. Meaningful coverage on business logic; we do not chase a percentage on getters.
- Architecture decision records. Every significant choice documented with its context and alternatives.
Delivery metrics we track
- 12×median increase in deployment frequency
- <30 minlead time from merge to production
- −71%reduction in change failure rate
How we choose
Technology selection happens in discovery and is written up with the alternatives considered, the trade-offs accepted and the conditions that would make us revisit the decision. You keep that document whether or not you keep us.
Applied
These stacks, in production
Case studies grouped by the kind of engineering they required.
SaaS
Orbit — real-time treasury platform
A multi-entity cash and liquidity platform replacing 40 spreadsheets and a nightly batch process with live positions across 14 currencies.
- 18×faster daily close
- 14currencies live-tracked
Mobile
Meridian — patient care companion app
An offline-first companion app for post-operative patients across 32 clinics, cutting readmissions through structured daily check-ins.
- −23%30-day readmissions
- 81%daily check-in compliance
E-commerce
Northwind — headless commerce replatform
A headless storefront over an existing commerce backend, rebuilt for speed and mobile conversion ahead of peak season.
- +31%mobile conversion rate
- 1.1slargest contentful paint
Available for new engagements
Not sure which stack fits?
That is a discovery question, and we answer it with evidence rather than preference. Tell us the constraints and we will tell you what we would choose.
- Reply within one business day
- No obligation, no sales sequence
- NDA on request before we talk