Senior by default
Every contributor on your project has shipped production systems for at least seven years. No staffing rotations, no junior shadow teams.
Volume 06 — Engineering Quarterly
Software written like it will outlive the people who shipped it.
TOOKI’S LIMITED is an engineering studio for organisations that have stopped trusting templates. We architect, build and operate the platforms that move money, regulate access, and translate ambition into infrastructure.
- Year founded
- 2019
- Engineers
- 84
- Active platforms
- 37
- Time zones
- 11
01Introduction
A firm of engineers, not vendors.
We were founded in Cardiff in 2019 by a small team that had spent a decade rebuilding other people’s systems after they failed in production. TOOKI’S LIMITED began as a deliberate response to a market that had grown accustomed to brittle integrations, opaque billing and vendor lock-in dressed up as innovation.
Today we operate as an independent engineering studio across the United Kingdom, the European Union and East Africa. We design distributed systems, build the products that run on them, and stay on as operating partners long after the launch press releases have gone quiet.
“We treat code as a long-term obligation, not a deliverable.”
02Why operators choose us
Six commitments that shape every engagement.
Written specifications
Every architecture decision is captured as a record. You inherit a system you can reason about long after we step away.
Operate what we build
We are accountable for the platforms we ship in production, not only in staging. Incident response is part of the contract.
Honest estimates
We size work in calendar weeks, not story points, and tell you when scope outgrows the budget instead of after.
Composable engagements
Engage one engineer or a full delivery team. Scale up or down monthly without contractual penalties.
Independent perspective
We hold no cloud partnerships, reseller margins or affiliate revenue. Our advice optimises for your bill, not ours.
03Values
Quiet craft, loud accountability.
- Restraint
- We add complexity only when the problem demands it. Fewer moving parts is always the first proposal.
- Legibility
- If a junior engineer cannot read the codebase in a week, the codebase is wrong, not the engineer.
- Stewardship
- We document, we hand over, we make ourselves replaceable. Loyalty is earned in renewals, not contracts.
- Honesty
- We name risk early, in writing, and treat budgets as commitments rather than aspirations.
04Expertise
The disciplines we have spent a decade refining.
Each engagement draws from these practice areas in combination. We do not staff projects from a single discipline in isolation.
- 01
Distributed backend systems
Event-driven architectures, durable queues, idempotent APIs and the schema-evolution discipline that keeps them running for years.
- 02
Front-end product engineering
React, TypeScript and design-system work where accessibility, performance and editorial polish are non-negotiable.
- 03
Data platforms & analytics
Warehousing on BigQuery, Snowflake and ClickHouse, with declarative pipelines and contract-tested transformations.
- 04
Applied machine learning
Retrieval-augmented generation, classical ML services, evaluation harnesses and the unglamorous plumbing in between.
- 05
Cloud and platform engineering
Hardened landing zones on AWS, GCP and Azure with terraform, policy-as-code and predictable cost behaviour.
- 06
Security & compliance
ISO 27001, SOC 2 Type II and PCI-DSS programmes treated as engineering work rather than paperwork.
05Industries
Where our work tends to land.
We bias toward industries where engineering decisions translate directly into measurable outcomes — uptime that affects revenue, latency that affects trust, audits that affect the licence to operate.
Financial infrastructure
Card schemes, ledgers and treasury systems for issuers, processors and neobanks.
Health & life sciences
Clinical platforms, regulated record systems, and the integrations that hospitals actually use.
Industrial & energy
SCADA-adjacent telemetry, asset management and forecasting platforms for utilities and operators.
Logistics & mobility
Routing, dispatch and customs systems that survive the volume spike of a Monday morning.
Public sector & policy
Service design and platform work for departments that need to deliver, not demonstrate.
Media & publishing
Editorial workflows, rights systems and high-traffic delivery for newsrooms and broadcasters.
06Technologies
The stack we reach for, and the one we avoid.
Languages
TypeScript, Go, Python, Rust, Kotlin and a small amount of Elixir where it earns its keep.
Runtimes
Node, Bun, Deno on the edge; containerised services on Kubernetes and Nomad in regulated environments.
Data
PostgreSQL, ClickHouse, Snowflake, BigQuery, DuckDB, Kafka, NATS and Temporal for orchestration.
Front-end
React, Solid, Svelte, design tokens managed in Style Dictionary, animation in GSAP and Motion.
Cloud
AWS, GCP and Azure landing zones built with Terraform, OpenTofu and Crossplane.
Observability
OpenTelemetry, Grafana, Honeycomb, Sentry and Prometheus, tuned to actually alert humans.
07Software development process
Eight stages, repeated until the system holds.
Discover
We interview operators, read the codebase you already have, and write down what we found.
Frame
We propose two to three architectural directions with trade-offs costed in time and cash.
Prototype
A narrow vertical slice ships in week three. It exercises the riskiest assumptions first.
Build
Two-week increments with working software at the end of each. Demos to the people who use it, not only those who buy it.
Harden
Load tests, chaos drills, threat modelling and a written runbook before anything goes near production.
Launch
We are on call for the first sixty days. Pager rotation, post-mortems, and weekly stability reviews.
Operate
Optional managed operations: 24/7 on-call, capacity planning, security patching and quarterly architecture reviews.
Hand over
Documentation, training, and a deliberate transition to your internal team or another partner.
08The team
Small enough to argue. Large enough to ship.
Eighty-four engineers, designers, and operators across three studios. Average tenure: six years. Median experience before joining: eleven. We hire deliberately and slowly, and we make a habit of refusing work that we cannot staff properly.
Cardiff studio
Headquarters, platform & security practices
Lisbon studio
Product engineering and design
Nairobi studio
Data platforms and applied ML
Distributed
Embedded engineers across nine further countries
09Innovation approach
Skepticism as a creative posture.
We treat novelty as a cost. New runtimes, frameworks and protocols enter our stack only after a deliberate evaluation: who maintains them, what they replace, and what happens to the system if they disappear in three years. The result is a portfolio of projects that ages well rather than impressively.
Our research practice ships a quarterly internal paper — currently on consensus replication, eventual consistency in regulated ledgers, and the practical edges of generative retrieval.
10Digital transformation
Modernisation, without the slogan.
We rebuild monoliths into composable platforms, retire vendor systems that have outlived their economics, and translate mainframe-era business logic into services your auditors and your engineers can both read.
The work is rarely glamorous. It is, however, the difference between an organisation that adapts and one that announces it intends to.
11Security standards
Defence in depth, written into the build pipeline.
Identity
Zero-trust access patterns, hardware-bound credentials, federated identity with mandatory step-up.
Code
Reproducible builds, signed artefacts, software bill of materials and dependency provenance.
Data
Field-level encryption, key rotation policies enforced by code, deterministic anonymisation for analytics.
Operations
Continuous control monitoring, evidence collection automated against ISO 27001 and SOC 2 controls.
12Infrastructure
Foundations that engineers can actually reason about.
We design landing zones the way a structural engineer designs a building: with deliberate boundaries, predictable failure modes, and the assumption that the people maintaining it in year seven will not be us.
Every account, project and namespace is created from a versioned blueprint. Every change is an audited pull request. Every region is paired with a recovery target measured in minutes, not aspirations.
13Cloud solutions
Hyperscale where it matters. Sovereign where it must.
We design hybrid topologies across AWS, GCP, Azure, OVH, Hetzner and on-premises hardware, with workload placement that reflects regulation, latency and economics rather than preference.
Public cloud
Account factories, FinOps governance, multi-region failover and predictable runtime cost.
Private cloud
Kubernetes and OpenStack platforms for regulated workloads with hardware-rooted attestation.
Hybrid edge
Edge compute pods deployed close to operations, orchestrated centrally, audited continuously.
14AI integration
Machine learning, treated like the rest of your platform.
Most of our machine-learning work is unglamorous. We build evaluation harnesses, label pipelines, retrieval indices and the model-routing layer that decides whether to use a fine-tuned classifier or a frontier model based on cost and accuracy.
The result is not a chatbot demo. It is a measured improvement to a workflow that someone in your organisation will perform ten thousand times this quarter, executed reliably enough that nobody notices the model behind it.
15Development methodology
An honest answer to ‘how do you work?’
We operate in two-week iterations with working software at the end of each. Every change is reviewed by at least one peer engineer. Every merge to main is followed by an automated deployment to staging within fifteen minutes.
Architectural decisions are captured as numbered records in the repository. Trade-offs are written down before they are made, not after they are regretted. The result is a system that can be onboarded by a new engineer in days rather than quarters.
16Case study
Re-platforming a European payments ledger from a vendor system to an owned, audited service.
Duration
14 months
Team
9 engineers
Outcome
−63% unit cost
A tier-one issuer migrated its core ledger off a hosted vendor onto an owned event-sourced service running across three regions. We designed the migration plan, built the platform, ran the dual-write reconciliation for nine months, and handed over a written runbook to the client’s in-house team at completion.
17Project workflow
The week, repeated.
- Monday. Joint planning between client and studio; commitments captured publicly.
- Tuesday — Thursday. Pair work, design reviews, two demo windows for stakeholders.
- Friday. Retrospective, written weekly note circulated to all sponsors before close of business.
18Quality assurance
Tests that mean something.
- Contract tests across every service boundary, run on every merge.
- Synthetic transactions that exercise live production paths and alert before customers notice.
- Quarterly red-team exercises against staging, with the findings driving the next quarter’s roadmap.
19Client experience
“They worked the way a senior engineering team works inside a company that takes itself seriously — written communication, honest estimates, no theatre. We renewed for a fourth year before the third one ended.”
20Statistics
99.987%
Median production uptime across managed platforms, trailing twelve months.
3.2m req/s
Peak sustained throughput on the largest platform we operate.
41
Active enterprise relationships across nineteen countries.
6 yrs
Average client tenure, weighted by contract value.
21Company timeline
Seven years, told plainly.
2019
Founded in Cardiff by four engineers with a shared distaste for vendor lock-in. First engagement: a payments ledger migration.
2020
Pandemic year. We doubled headcount, opened the Lisbon studio, and shipped two regulated platforms remotely.
2021
ISO 27001 certified. First public-sector engagement. Built our internal platform-as-code base now used in every project.
2022
Opened the Nairobi studio. Began a multi-year programme with a tier-one European card issuer.
2023
SOC 2 Type II report issued. Crossed thirty active enterprise relationships.
2024
Published our internal handbook externally. Launched the applied-ML practice with three founding engineers.
2025
Crossed eighty engineers. Began a multi-region replatforming for a national health service supplier.
2026
Operating fifty-plus production platforms. Quietly profitable, deliberately small, still entirely independent.
22Business advantages
What you measurably get.
Predictable runtime cost
FinOps governance baked into our landing zones tends to surface 20–40% reductions in the first quarter.
Reduced vendor exposure
Our default is to give you the system, not a managed dependency on us.
Audit-ready by construction
Evidence for ISO 27001, SOC 2 and PCI-DSS controls is collected automatically as the platform runs.
Talent retention
Teams that work alongside our engineers tend to keep their senior staff. We optimise for hand-over from day one.
23Frequently asked
Twelve questions we hear most often.
How small an engagement will you take?
We routinely embed a single senior engineer for three-month engagements alongside in-house teams.
Do you work on a fixed price?
Occasionally, for narrow well-defined scopes. Most of our work is time-and-materials with a written ceiling.
Do you use offshore subcontractors?
Never. Every engineer is on staff, salaried, and named on the engagement.
How do you handle IP?
All work product is assigned to the client on payment, with a clear carve-out for our internal libraries and tooling.
What time zones do you cover?
Our studios overlap from 06:00 UTC through 22:00 UTC. Off-hours on-call is contracted explicitly.
Do you sign NDAs?
Routinely. We also publish a standard mutual NDA template if you prefer to skip the negotiation.
Who owns the cloud accounts?
You do. We deploy into accounts under your control and never the other way around.
Can you take over an existing codebase?
Yes. Roughly half of our engagements begin with another team’s code. We write a discovery report in the first three weeks.
Do you require minimum contract length?
Three months for engagements over two engineers, monthly thereafter.
How do you price applied ML work?
We price evaluation and platform work separately from model experimentation. You pay only for the work that has shipped.
Do you offer training?
Yes — structured workshops on platform engineering, applied ML and incident response, delivered by the engineers who do the work.
Are you hiring?
Continuously, and slowly. We close a search only when the candidate raises the average.
24Future vision
The next decade is operational, not visionary.
The interesting work of the coming decade will not be the announcement of new platforms. It will be the deliberate, unglamorous integration of the platforms we already have — with stronger guarantees about cost, latency, identity and the obligation to explain ourselves to a regulator.
Our roadmap is shaped accordingly. We continue to invest in platform engineering, applied ML evaluation, and the boring, high-leverage work of making systems legible to the people who depend on them.
