Setusphere

Engineering capabilities

Engineering Capabilities Built for Complex Workflows.

Four technical pillars, each covering the decisions that determine whether a system is still workable in year three. Every pillar links to the practice pages behind it — including the parts we will tell you to skip.

Schedule Technical AuditBrowse architecture blueprints

Technical pillars

Custom SaaS & Web Platforms

Multi-tenant products designed to survive their own growth.

The decisions that determine whether a platform is still workable in year three are all made in the first month: how tenants are separated, how permissions are described, and how the outside world is allowed to talk to it. We settle those on paper before they become migrations.

Multi-tenant isolation
An isolation strategy chosen deliberately — shared schema, schema-per-tenant or database-per-tenant — and one you can defend to a security reviewer.
RBAC schemas
Permissions modelled as data rather than conditionals, so a new role is a row and not a release.
Dynamic API hooks
Versioned endpoints and outbound webhooks with retries and a delivery record, so integrations fail loudly rather than silently.
SaaS Product EngineeringWeb ApplicationsMobile Applications

Enterprise Automations & ERPs

Algorithmic engines that replace a recurring manual process.

Most enterprise back-office work is a constraint problem being solved by hand every morning — cover allocation, shift rostering, fee reconciliation, approval routing. These are solvable, but only if the rules are written down and the result stays overridable by the person accountable for it.

Constraint solvers
Allocation and scheduling with the rule visible in the interface, weighted for fairness over a period rather than optimised for a single day.
Workflow engines
Approval chains, state machines and escalation paths with an audit trail, because 'who approved this' gets asked later at the worst time.
Custom HR & finance logic
Ledgers, payroll rules and reconciliation that match your actual policy instead of the nearest option on a settings screen.
Architecture & Technical Review

Applied AI & Machine Learning

Model features shipped behind an evaluation harness, not a demo.

An AI feature is easy to demonstrate and hard to trust. The engineering that separates the two is unglamorous: a retrieval layer that cites its source, an evaluation suite that runs on every change, and a deterministic path for when the model is not confident enough.

RAG pipelines
Chunking, embedding and retrieval over your own corpus, with every answer traceable to the document it came from.
Document extraction agents
Structured fields pulled out of invoices, forms and contracts, with a human review queue for anything below the confidence threshold.
Predictive analytics
Forecasting and scoring against your historical data, reported with its error bars rather than as a single confident number.
Data & Applied AI

Cloud Engineering & Infrastructure

Environments recreated from source, deployments that stay boring.

Infrastructure earns its keep on the worst day, not the average one. The test is whether you can rebuild an environment from source without a person remembering a manual step, and whether a release can go out at 3pm without anyone holding their breath.

DevOps pipelines
CI/CD producing the same artefact twice, promoted through environments rather than rebuilt for each one.
Auto-scaling microservices
Container orchestration with resource limits and cost visibility, so scaling is a configuration decision rather than an invoice surprise.
Zero-downtime migrations
Expand-and-contract schema changes and blue/green releases, with a tested rollback path — not an assumed one.
Platform & Cloud Infrastructure

Engagement models

Three shapes, compared side by side

Which one applies is usually obvious once the scope is written down. We would rather settle it then than renegotiate it in month three.

Fixed scope

One deliverable, one price

Best for
Work we can specify precisely — a site, an integration, a first release.
Commitment
The agreed deliverable, nothing open-ended.
Billing
Fixed price, staged against milestones.
Scope change
A written change to the scope and the price. Nothing absorbed quietly.
Typical length
4–20 weeks

Embedded team

Engineers inside your process

Best for
Continuous product development against a roadmap that keeps moving.
Commitment
Monthly, one month's notice.
Billing
Per engineer, per month.
Scope change
Reprioritise at any sprint boundary. That is the point of this model.
Typical length
3 months minimum

Retainer

The maintenance nobody budgets for

Best for
A live system that needs upgrades, incidents and small changes handled.
Commitment
An agreed block of hours each month.
Billing
Monthly block. Unused hours do not roll forever.
Scope change
Anything inside the block. Larger work is quoted separately.
Typical length
Rolling, monthly

Process

How a project runs

The stages are identical for a multi-tenant platform and a marketing site — only their length changes. Select a stage to see what physically exists when it ends.

Stage 01

Discovery & Blueprint

A working session to fix the problem, the users and the cut list. You leave with a written scope rather than a proposal deck — including the features we recommend not building.

What exists at the end

  • Written scope and cut list
  • Success criteria
  • Risk register

Not sure which pillar you need?

Most projects touch two or three. Describe the problem and we will tell you which parts actually apply — including the ones you can skip.

Book a 15-Min Strategy Call

We reply within one working day.

Chat on WhatsApp