01Practice
Custom Software Development
Most companies reach us with a spreadsheet that outgrew itself, or a vendor product that fits eighty percent of the process and fights the rest. We build the twenty percent that is actually your business — typed, tested, documented, and owned by you. No licence renewal, no roadmap you cannot influence.
- Typical duration
- 2–3 weeks to start · 3–5 weeks
- Engagement models
- Scoped build · Embedded pod · Discovery first
- Practice lead
- Named at proposal stage

01The problem
The gap between the process and the product
Every company runs a handful of processes that no packaged product models correctly. Claims triage. Freight consolidation. Clinical intake. The work-arounds start small — an export here, a shared sheet there — and within two years the real system of record is a folder nobody has permission to delete. Finance reconciles by hand. Onboarding a new hire takes a week of tribal knowledge.
The usual answers both fail in predictable ways. Bending a vendor product past its intended shape produces a customisation bill that renews annually and blocks upgrades. Hiring a permanent team to build it in-house is the right answer eventually, but not while the problem is still being defined. We sit in between: build the system properly, prove it in production, then hand it to the people who will live with it.
02What GatiQ delivers
The work itself
Every item below is something you can inspect, run or read at the end of the engagement.
- 01
A domain model you can argue with
Before any framework is chosen, we write the entities, the state transitions and the invariants down in plain language and walk your operations lead through them. Disagreements surface in week two, on a whiteboard, instead of in month six inside a database migration.
- 02
One system of record, with the edges handled
Idempotent write paths, explicit retry and compensation for anything that crosses a network, and a reconciliation job that proves the ledger balances. The unhappy paths are specified with the same care as the happy one, because they are where the finance team lives.
- 03
A migration that runs twice before it runs for real
We rehearse the cutover against a production-shaped copy, measure how long it takes, and write the rollback before we write the go-ahead. Dual-write or shadow-read periods keep the old system authoritative until the numbers match on both sides.
- 04
Tests that survive a refactor
Fast unit coverage on the domain logic, contract tests at every service boundary, and a small suite of end-to-end journeys in Playwright that mirror what your busiest user does before lunch. They run on every pull request and gate the merge.
- 05
Documentation aimed at your next engineer
A README that gets a laptop running in under thirty minutes, decision records listing the options we rejected and why, and a diagram generated from the deployment manifests so it cannot drift away from reality.
03Engagement model
Three ways to buy it
- Scoped build
- Fixed scope, a date range rather than a date, and a weekly demo. Best when the outcome is well understood and the deadline is external — a regulatory change, a contract start, a warehouse opening.
- Embedded pod
- Two to five GatiQ engineers inside your existing team, using your board, your standup and your code review. Monthly rate, thirty days notice either way. Best when you have direction but not enough hands.
- Discovery first
- A short, fixed-fee engagement ending in a written architecture and a costed plan. It is yours to keep — you can take it to another vendor without penalty, and some clients do exactly that.
04Typical timeline
What the calendar looks like
- 01
Discovery
2–3 weeksProcess interviews, code and data audit, written architecture.
- 02
Foundation
3–4 weeksDomain model, auth, CI, environments, first vertical slice in production.
- 03
Build
10–20 weeksTwo-week increments against an agreed backlog, demoed from staging.
- 04
Cutover and handover
3–5 weeksRehearsed migration, runbooks, pairing with your team, on-call handoff.
Ranges, not dates. We narrow them at the end of discovery and tell you the same week if they move.
05Technologies
What we reach for
Defaults, not rules. If your team already runs something else competently, that is usually the better choice.
- Languages
- TypeScript, Go, Python, Kotlin
- Server
- Node.js, NestJS, FastAPI, gRPC, GraphQL
- Data
- PostgreSQL, Redis, ClickHouse, Kafka
- Workflow
- Temporal, Celery, River
- Verification
- Vitest, Playwright, Testcontainers, k6