Hire elite Go developers.
High-performance backend, cloud-native.

Expected answer

Goroutines are lightweight green threads managed by the Go runtime — they start with ~2KB stack (vs. 1MB for OS threads) and are multiplexed onto OS threads by the scheduler. Channels provide typed, synchronized communication between goroutines. Unlike shared memory with mutexes, channels follow the CSP model: "Don't communicate by sharing memory; share memory by communicating." This eliminates most race conditions by design.

Expected answer

Use a token bucket algorithm with Go's time.Ticker and channels. For distributed rate limiting: Redis-based sliding window with Lua scripts for atomicity. The Go standard library's rate.Limiter is a good starting point for single-instance limiting. Key design decisions: per-client vs. global limits, burst allowance, graceful degradation (return 429 with Retry-After header), and metrics for monitoring limit hits.

Expected answer

Arrays have fixed size at compile time (e.g., [5]int), while slices are dynamic views backed by arrays ([]int). Slices have a length, capacity, and pointer to the underlying array. Appending beyond capacity allocates a new, larger array and copies data. Key gotcha: multiple slices can share the same backing array — mutating one affects others. When in doubt, use copy() for independent slices.

Expected answer

Go uses explicit error returns instead of exceptions. Patterns: 1) Wrap errors with context (fmt.Errorf with %w for wrapping), 2) Use errors.Is() and errors.As() for checking, 3) Sentinel errors for expected cases (e.g., ErrNotFound), 4) Custom error types with structured fields for rich context, 5) Never ignore errors (linter enforcement with errcheck). For production: structured logging with the error chain, proper error boundaries at API handlers that map internal errors to HTTP responses.

Expected answer

Standard layout: cmd/ (entry points), internal/ (private packages), pkg/ (public packages if any). Use dependency injection (no frameworks needed, just interfaces). Configuration: env vars with a config struct. HTTP: Chi or standard mux + middleware chain (logging, auth, CORS, rate limiting). Database: pgx for PostgreSQL with connection pooling. Testing: table-driven tests, testcontainers for integration. Observability: OpenTelemetry for traces, Prometheus for metrics, structured logging (slog). Deployment: multi-stage Docker build for minimal image.