I Hired 3 Solutions Architects in Singapore and Only 1 Stuck — the 7-Step Filter I Use Now

The second one told me exactly why he was leaving, in a sentence I have thought about ever since: “I have been asked to be accountable for decisions I am not allowed to make.” He had spent eleven months writing design documents that three teams treated as suggestions. He was not wrong, and the failure was mine and it was structural.

What follows is the method that came out of that, refined across nine solutions architect searches in Singapore. It assumes you are hiring your first or second architect rather than adding to an established practice, and it is specific to the Singapore market on salary, pass processing and pool composition.

Step 1 — Decide whether you need an architect or a senior engineer

This decision precedes everything and getting it wrong wastes the full cycle plus a good hire.

A senior engineer owns a system deeply. Its code, its performance, its failure modes, its evolution. Success is measured in how well that system works.

A solutions architect owns the space between systems. Integration, boundaries, three-year commitments, and trade-offs whose constraints are frequently commercial rather than technical. Success is measured in how well the estate holds together.

The test that settles it is scope of consequence. If the hardest decisions the role faces sit inside one codebase, you want a senior or staff engineer. Hiring an architect into that mandate produces a frustrated hire who leaves inside a year. If the hardest decisions span three teams, two vendors and a regulator, that is architecture.

This matters commercially in Singapore: architect-titled candidates carry a 15 to 25% premium over equivalent senior engineers. Paying it for work that is actually deep single-system engineering is a double loss.

Step 2 — Write the job description around decision rights

This is the step I skipped on the two hires that failed, and it is the one I would keep if I could keep only one.

Most architect job descriptions describe activities: design reviews, technology evaluation, standards, stakeholder alignment. Every architect does those things everywhere. What differs between organisations, enormously, is what happens when the architect and a delivery team disagree.

Write three lists, explicitly, in the job description:

• Decisions this role owns. Made by the architect, teams implement. Be concrete: data residency approach, inter-service communication patterns, choice of managed service for a given capability.
• Decisions this role advises on. Team decides, architect is consulted and can escalate.
• Who can overrule, and by what process. Somebody can. Name them and name the mechanism.

Strong candidates read these three lists first and ask sharper follow-ups than about salary. Candidates who do not ask about decision rights at all are telling you something useful.

Step 3 — Source on delivered systems, not certifications

Searching Singapore for architect titles returns a pool weighted heavily toward large systems integrators and consultancies, where the job frequently means producing documents rather than living with their consequences. That pool is not bad, but it is not representative, and it is where our two failed hires came from.

The signal we source on now is having run a system through a full lifecycle including its failures. Concretely: built it, operated it, watched it break, and either fixed or replaced it. Fourteen of the twenty-seven credible candidates across our searches had staff or principal engineer titles rather than architect titles.

On certifications, the data from our own loops is blunt. Across twenty-seven candidates, certification status had effectively no relationship with design review performance. Certifications test recall of a vendor's recommended patterns — useful for a delivery engineer, close to irrelevant for someone whose job is deciding when the recommended pattern is wrong for your constraints. Tiebreaker only.

Geographically, Singapore architect searches draw from three pools: candidates already in Singapore on valid passes (fastest, most competitive), regional candidates from Malaysia, India, Indonesia and the Philippines, and returning Singaporeans from London, Sydney or the US west coast. The third group is small and consistently strong, and it is the one most employers forget to target directly.

Step 4 — Screen on trade-offs the candidate got wrong

One question, twenty-five minutes, and it changed our hit rate more than anything else:

“Describe an architecture decision you made that you later reversed. What did the reversal cost, and what would you have needed to know at the time?”

Four things to listen for.

Do they have one at all? An architect who has never reversed a decision has never owned one. This alone eliminated six of twenty-seven.

Do they know the cost? Strong candidates quantify: engineer-months, a delayed launch, a migration that consumed two quarters. Vague answers indicate distance from consequences.

Was the missing information knowable? The best answers identify something that could have been discovered cheaply and was not — a load test not run, a regulator not consulted, a team not asked. Weak answers blame circumstances that changed.

Do they describe the human cost? Reversals mean telling engineers their work is being discarded. Candidates who mention this have led through it. Candidates who describe reversals purely technically have watched them from a distance.

Step 5 — Run a 90-minute design review on a real system

Drop the whiteboard greenfield exercise. “Design a URL shortener” tests pattern recall and rewards candidates who have rehearsed, which is the opposite of what you need.

Instead, take one of your real designs — lightly anonymised, with its actual constraints and its actual compromises — and ask the candidate to review it. Give it to them thirty minutes ahead.

What we grade:

Do they ask about constraints before criticising? Every real design contains decisions that look wrong without context. Candidates who critique immediately will do the same thing to your teams in week two.

Do they distinguish “wrong” from “not how I would do it”? This is the single most important architect trait and it is directly observable here. An architect who cannot tell the difference generates enormous friction and very little value.

Do they prioritise? Strong candidates name the two things that matter and explicitly set aside the rest. Weak candidates produce twenty observations of equal weight.

Have two engineers who work on that system in the room. Their read is usually better than yours, and if the architect is hired they have already had a substantive technical conversation with the team.

Step 6 — Benchmark against 2026 Singapore bands

Financial services and regulated sectors pay at the top of each band and sometimes above, because the regulatory dimension narrows the pool sharply. Expect a variable component of 10 to 20% at senior levels.

One Singapore-specific warning: this market reprices scarce architecture skills faster than any other regional hub. A band you benchmarked twelve months ago is likely materially out of date, and candidates know it. Benchmark fresh before every search, not annually.

Step 7 — Handle pass timing and write the 90-day mandate

Plan 55 to 85 days from requisition to first day. Sourcing four to seven weeks, interviewing two to three, notice periods commonly two to three months at this level and rarely reduced.

For candidates already in Singapore on a valid pass, transfer processing is short and predictable. For overseas candidates, start Employment Pass processing under COMPASS at offer acceptance so it runs in parallel with notice. Architect-level salaries usually clear the qualifying criterion comfortably, but the MyCareersFuture advertising requirement still applies unless an exemption is available, and forgetting it costs two weeks.

Then write the 90-day mandate. Ours has three lines: what they will have documented by day 30 — usually the current state, honestly — which single decision they will have made by day 60, and what measurable outcome exists by day 90.

Keep the day-60 decision small and real. An architect who has made one genuine decision and had it implemented is established. An architect who has produced a target-state document is not, however good the document is. That distinction is the whole difference between our one success and our two failures.

How this compares across the region

Architecture hiring differs sharply by hub.

See also our guides to hiring a data engineer in Singapore and hiring DevOps engineers for GPU cloud, and our analysis of what the regional AI investment split means for 2027 headcount.

Frequently Asked Questions