Internal R&DIn progress
Project one
A first internal project write-up is being prepared: the problem, the architecture, the trade-offs and what we would do differently.
- Architecture
- TypeScript
Software engineering studio
We build the softwarethat other softwaredepends on.
Arena Dream Factory designs and engineers sophisticated systems — architecture, algorithms, real-time data, developer tooling and interfaces that hold up under load. Small team, senior hands, no layers between you and the people writing the code.
- < 24h
- Reply to every inquiry
- 100%
- Senior engineers on delivery
- Day 1
- You get access to the repo
01
Precision
Requirements get written down before code gets written. Edge cases are named, not discovered in production. Every decision that costs money later is made deliberately now.
02
Depth
Complex domains are where we are useful — concurrency, state machines, protocol work, performance budgets, data models that have to survive five years of change.
03
Communication
Weekly written updates, a shared board you can read at any hour, and a straight answer when something slips. You never have to ask what is going on.
04
Ownership
We hand over documented, tested, buildable code and a runbook. No lock-in, no undocumented magic, no dependency on us to keep the lights on.
Capabilities
Six things we do properly, instead of twenty we do adequately.
Engagements usually combine two or three of these. If what you need is not on the list, say so anyway — a straight no costs you one email.
Product engineering
End-to-end delivery of web products — from a blank repository to something running in production with monitoring, CI and a deployment story.
- Full-stack web applications (TypeScript, React, Astro, Node)
- API design — REST, GraphQL, typed RPC contracts
- Auth, billing, multi-tenancy and permission models
Systems & architecture
The structural work: how the pieces fit, where the state lives, what happens when a node dies and what the thing costs to run at ten times the traffic.
- Domain modelling and service decomposition
- Event-driven pipelines, queues and workflow orchestration
- Data layer design — Postgres, SQLite, object storage, caching
AI & automation
Language models wired into real workflows, with evaluation and guardrails — not a chat box bolted onto a landing page.
- Retrieval pipelines over your own documents and data
- Agentic workflows and tool integrations
- Evaluation harnesses so quality is measured, not assumed
Interface craft
Interfaces built by people who also write the backend, so what you see is fast, accessible, and honest about what the system is doing.
- Design systems and component libraries
- Data-dense dashboards, editors and internal tools
- Accessibility (WCAG 2.2 AA) and keyboard-first flows
Rescue & review
Someone else built it, it is slow, fragile or nobody understands it any more. We read it, map it, and give you options with numbers attached.
- Code and architecture audits with a written report
- Performance and cost investigations
- Incremental modernisation instead of a risky rewrite
Engineering partnership
Ongoing capacity for teams that need senior hands without another full-time hire — embedded in your process, your repo, your standups.
- Monthly retainer with a fixed, predictable envelope
- Code review and mentoring for your in-house team
- On-call handover, documentation and runbooks
How it goes
Five stages. You see working software after the second one.
Nothing here is a black box. Every stage produces an artefact you can read, review or take somewhere else.
- 01
Frame
Days 1–3 · free
A call and a written brief. We define what success looks like, what is explicitly out of scope, and what the riskiest unknown is. If we are the wrong studio for the job, this is where we say so.
- 02
Shape
Week 1–2
Architecture sketch, data model, interface flows and a cost estimate with the assumptions attached. You get a document you could hand to any other engineer — it is yours either way.
- 03
Build
Two-week cycles
Working software at the end of every cycle, deployed to a preview environment you can click through. Written update each Friday: what shipped, what moved, what is at risk.
- 04
Harden
Before every release
Tests, load checks, accessibility pass, error tracking, and a rollback plan. A release is not done when the code merges — it is done when it survives real traffic.
- 05
Hand over
Ongoing
Documentation, runbook, architecture notes and a walkthrough with your team. You keep the repository, the infrastructure accounts and the knowledge. Support continues only if you want it to.
Toolkit
Boring technology, used well.
We pick tools for how they behave in year three, not how they demo in week one. The list changes slowly and on purpose.
Languages
- TypeScript
- Python
- Rust
- SQL
- Go
Frontend
- React
- Astro
- Svelte
- Tailwind
- Web Components
Backend
- Node
- FastAPI
- PostgreSQL
- SQLite / D1
- Redis
Platform
- Cloudflare Workers
- AWS
- Docker
- Terraform
- GitHub Actions
AI
- Claude
- Embeddings & vector search
- RAG pipelines
- Eval harnesses
Practice
- Trunk-based CI
- Typed contracts
- Observability
- Threat modelling
Selected work
Internal projects, published as we go.
Arena Dream Factory is a young studio. Rather than dress up a client list we do not have yet, we are publishing our own projects — the tools and systems we build to sharpen the practice. Each one ships with its architecture notes.
Internal R&DIn progress
Project two
Second study in the queue — including benchmarks and the measurements behind the design decisions.
- Performance
- Data
Internal R&DIn progress
Project three
Third study in the queue, focused on interface craft and how the system behaves at the edges.
- Interface
- AI
“Sophistication is not the number of moving parts. It is knowing exactly which parts have to move, and being able to explain why to the person paying for them.”
Next step
One call, one written brief, no invoice.
If what you need is clear, we will shape it. If it is not clear yet, that is exactly what the first conversation is for.