Solution Architect Interview Questions

Option B is the strongest: it grounds the design in specific, quantified business requirements (sub-100ms, €2B volume, 99.99%), walks through the architecture by layer with the reasoning behind each choice, articulates real trade-offs (accuracy vs. latency, consistency vs. SLA), and demonstrates that the architect considered and resolved conflicting constraints — which is what makes a design "complex" in the interviewer's eyes. How to structure a system design walk-through in an interview (the PAR+D pattern): Problem — state the specific constraints that made it challenging. Numbers matter: "millions of requests" is vague; "sub-100ms at p99" is specific. Architecture — walk through the system by layer or by request flow. Name every component and why you chose it (not just what it is). Rationale for choices — "I chose Kafka over an HTTP-based queue because of its replay capability and the fan-out to multiple consumer services." Trade-offs acknowledged — the mark of a senior architect is knowing what they gave up. "This approach sacrifices strong consistency for availability — here's why that was acceptable." Key architectural vocabulary: Latency tiers — fast path vs. slow path. Feature store — a system for serving pre-computed ML features at low latency. Tiered architecture — separating concerns by latency requirements. PCI-DSS — Payment Card Industry Data Security Standard. gRPC — Google's high-performance RPC framework (binary protocol, lower overhead than REST/JSON).

Option B is the strongest: it reframes the conflict as a technical trade-off signal, introduces the specific facilitation technique of separating interests from positions, gives a concrete comparison-matrix approach, demands quantified acceptance criteria (which is the architect's job), takes ownership of the recommendation ("I say..."), and explains the value of a decision log. The "interests vs. positions" framework comes from negotiation theory (Fisher & Ury, "Getting to Yes") and is highly relevant to architecture work. Positions are specific demands ("we need real-time data"). Interests are the underlying needs ("we need to make decisions quickly enough to retain customers"). Key architecture communication vocabulary: Trade-off matrix — structured comparison of options across multiple dimensions (cost, latency, consistency, complexity). Acceptance criteria — specific, measurable, testable thresholds that define when a requirement is met. "Fast" → "p99 response time under 200ms." Decision log / Architecture Decision Record (ADR) — a document recording: the decision made, the context at the time, the alternatives considered, the rationale, and the consequences. Prevents relitigating settled decisions when team members change. Own the recommendation — solution architects are expected to make recommendations, not just present options. "It depends" is acceptable if followed by "and here is what it depends on, and here is my recommendation given our context." Options C and D are solid but less structured and miss the "interests vs. positions" framework.

Option B is the strongest: it names the four specific dimensions executives evaluate (cost, time, risk, capability), works through a complete concrete example with real numbers (making it actionable rather than abstract), frames the time risk of deferral (forced emergency migration), makes a recommendation, and prepares for one of the most common executive follow-up questions ("why didn't we do this earlier?") — a detail that shows presentation experience. The four executive decision dimensions: Cost — CapEx (capital: servers, licences) vs. OpEx (ongoing: team time, cloud spend). Quantify in money if possible. Time — time to market, implementation timeline, opportunity cost of delay. Risk — operational risk (what breaks?), compliance risk (GDPR, SOC 2?), strategic risk (vendor lock-in?). For each option, name both the risk of choosing it AND the risk of not choosing it. Capability — what can the business do after this decision that it couldn't before? What does it no longer be able to do? This is what makes technical decisions strategic. The "risk of inaction" framing: executives often underestimate the cost of not acting. "Option A has zero implementation risk, but carries an 18-month scalability risk. Option B removes the 18-month risk." This reframes the trade-off as a risk decision, not a spending decision.

Option B is the strongest: it opens with the precise framing principle (differentiated value vs. commodity), gives four structured evaluation dimensions with specific anti-patterns (build cost typically underestimated 3×, security liabilities of self-built auth), names a concrete decision heuristic ("would a competitor see this as a differentiator?"), and includes the regulated industry exception — showing real-world nuance. Build vs. buy decision framework for architecture interviews: The core question is always: is this where we win, or is this infrastructure? Differentiating capabilities → build or build on OSS. Commodity capabilities → buy or use SaaS. Common examples: Buy: payments (Stripe), auth (Auth0, Cognito), email delivery (SendGrid), monitoring (Datadog), search (Elastic Cloud). Build (often): recommendation engine, core ML model, proprietary matching algorithm, custom reporting, domain-specific workflow. The underestimated build costs: Engineering time × 3× — initial estimates are almost always wrong. Maintenance burden — someone owns it forever: bug fixes, security patches, feature requests. Operational cost — infrastructure, monitoring, on-call. Opportunity cost — every engineer-month on internal tooling is an engineer-month not building product. Vendor lock-in mitigation strategy: wrap third-party APIs behind an internal abstraction layer so you can swap vendors without changing business logic. This is the Adapter pattern applied at the service level.

Option B is the strongest: it gives a complete taxonomy of practices (ADRs, fitness functions, dependency management, RFC process, evolutionary architecture), explains the specific problem each practice solves (ADRs prevent relitigating decisions, fitness functions prevent drift, dependency tracking detects decay), and introduces the concept of "evolutionary architecture" — which signals a senior architect's thinking about designing for change rather than predicting the future. Key architectural maintainability vocabulary: Architecture Decision Record (ADR) — a short document recording: status (proposed/accepted/deprecated), context (what forces are at play?), decision (what we decided), consequences (what becomes easier/harder). Michael Nygard's format is standard. ADR repository: a /docs/decisions/ folder in the repo, versioned with the code. Fitness function — from "Building Evolutionary Architectures" (Ford, Parsons, Kua). An automated test that verifies an architectural property holds. Examples: no module-level cyclic dependencies; all API responses include correlation IDs; no synchronous calls from service A to service B. These run in CI and prevent architectural drift without relying on code review. Evolutionary architecture — the practice of designing systems with change as a first-class concern. Loose coupling means a change in one component requires minimal changes elsewhere. Defined contracts (OpenAPI, Protobuf) enable independent evolution. Hexagonal architecture / ports and adapters isolates the core domain from infrastructure. Request for Comments (RFC) process — a lightweight written review for significant technical decisions. Used at Stripe, Uber, Netflix. More efficient than meetings for complex decisions; creates a searchable decision history.