Architecture scenarioMedium~12 minAI Application Architecture / 3 of 18

Interview question

How would you encapsulate a model provider inside an application?

Creates a task-oriented model boundary that normalizes application outcomes without erasing useful provider capabilities.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

I keep provider SDK types behind an infrastructure adapter and expose task-oriented application operations such as classify, extract, or generate a supported answer. The boundary accepts application inputs and returns typed outcomes including success, refusal, incomplete output, validation failure, throttling, and timeout. Provider request IDs, usage, and finish details remain available for diagnostics without leaking through every caller.

Full answer below

Practice focus

Interface design, provider error mapping, feature exposure, and maintainable integration boundaries.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Overview
A provider SDK is an infrastructure detail, but a universal lowest-common-denominator interface is also a trap.
I keep provider SDK types behind an infrastructure adapter and expose task-oriented application operations such as classify, extract, or generate a supported answer. The boundary accepts application inputs and returns typed outcomes including success, refusal, incomplete output, validation failure, throttling, and timeout. Provider request IDs, usage, and finish details remain available for diagnostics without leaking through every caller.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
Shared concerns such as credentials, model identifiers, timeouts, retries, tracing, and cost accounting live in the adapter or a deliberate gateway. Task-specific configuration stays near the use case instead of becoming one giant generic generate method with dozens of optional parameters. Contract tests use captured or provider test responses to verify error mapping and streaming behavior.
Finish
Advanced answer
Optional deeper angle for production judgment and nuance.
I do not pretend all providers support identical semantics. A small capability model or provider-specific extension can expose features such as schema-constrained output, caching, or reasoning controls without contaminating unrelated use cases. The application chooses behavior based on declared capabilities and tested policy, not string checks scattered through code. Provider upgrades are canaried behind versioned adapters, and raw responses are retained only when privacy and incident needs justify them.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Returning provider response objects from domain services.
Creating one generic interface with every provider option.
Discarding request IDs and usage metadata.

Red flagsCandidateRed flagsSignals that an interviewer may treat as risky.Open details

Details

Changing providers requires edits across controllers and business logic.
Every exception becomes a generic internal error.
Provider capability checks are string comparisons throughout the codebase.

Follow-up prompts

Quick interviewer prompts inside this drill.

Follow-up questionsCandidateFollow-up questionsLikely next questions after the main answer.Open details

Details

Treat these as the next interviewer turn. Answer one out loud, then save the gap in your practice notes.

What belongs in a task-oriented interface?
How would you expose provider-specific features?
Which errors need distinct application outcomes?

Continue the section

Next in AI Application Architecture: When to use an AI gateway

Step 4 of 18

When would you put a gateway in front of model providers?

Practice route

Read core answer Review checklist Follow-up prompts Track progress

Finished this drill?

Practice freely. Sign in only when you want to save completion, confidence, notes, or bookmarks.

New to Aporeon? Create account

Drill details

Human editorial direction

Aporeon is shaped by Aleksandar Tomovski, a software developer with experience on both sides of technical interviews. Content is reviewed for accuracy, natural spoken delivery, useful depth, and honest trade-offs.

About the creator Editorial standards

Was this drill useful?

Help improve the interview library.

Found this answer useful?

Say thanks with a standalone, one-time $5 contribution. No account required.

No subscription or perks.Choose another amount

Opening drill

Architecture scenarioMedium~12 minAI Application Architecture / 3 of 18

Interview question

How would you encapsulate a model provider inside an application?

Creates a task-oriented model boundary that normalizes application outcomes without erasing useful provider capabilities.

1Say itAnswer out loud first

2CompareCompare with Core answer

3RefineRefine with Stronger and Advanced

TL;DR

Full answer below

Practice focus

Interface design, provider error mapping, feature exposure, and maintainable integration boundaries.

Answer first

Answer path

Core answer first, then checkpoints toward a stronger answer.

Start
Core answer
Start here. This is the direct spoken answer to practice first.
Overview
A provider SDK is an infrastructure detail, but a universal lowest-common-denominator interface is also a trap.
I keep provider SDK types behind an infrastructure adapter and expose task-oriented application operations such as classify, extract, or generate a supported answer. The boundary accepts application inputs and returns typed outcomes including success, refusal, incomplete output, validation failure, throttling, and timeout. Provider request IDs, usage, and finish details remain available for diagnostics without leaking through every caller.
Checkpoint
Stronger answer
Optional upgrade with examples, trade-offs, and sharper phrasing.
Shared concerns such as credentials, model identifiers, timeouts, retries, tracing, and cost accounting live in the adapter or a deliberate gateway. Task-specific configuration stays near the use case instead of becoming one giant generic generate method with dozens of optional parameters. Contract tests use captured or provider test responses to verify error mapping and streaming behavior.
Finish
Advanced answer
Optional deeper angle for production judgment and nuance.
I do not pretend all providers support identical semantics. A small capability model or provider-specific extension can expose features such as schema-constrained output, caching, or reasoning controls without contaminating unrelated use cases. The application chooses behavior based on declared capabilities and tested policy, not string checks scattered through code. Provider upgrades are canaried behind versioned adapters, and raw responses are retained only when privacy and incident needs justify them.

Review signals

Use these to catch weak phrasing, shallow reasoning, and risky claims.

Common mistakesCandidateCommon mistakesMistakes that weaken the answer or signal shallow understanding.Open details

Details

Returning provider response objects from domain services.
Creating one generic interface with every provider option.
Discarding request IDs and usage metadata.

Red flagsCandidateRed flagsSignals that an interviewer may treat as risky.Open details

Details

Changing providers requires edits across controllers and business logic.
Every exception becomes a generic internal error.
Provider capability checks are string comparisons throughout the codebase.

Follow-up prompts

Quick interviewer prompts inside this drill.

Follow-up questionsCandidateFollow-up questionsLikely next questions after the main answer.Open details

Details

Treat these as the next interviewer turn. Answer one out loud, then save the gap in your practice notes.

What belongs in a task-oriented interface?
How would you expose provider-specific features?
Which errors need distinct application outcomes?

Continue the section

Next in AI Application Architecture: When to use an AI gateway

Step 4 of 18

When would you put a gateway in front of model providers?

How would you encapsulate a model provider inside an application?

Practice focus

Answer path

Overview

Review signals

Follow-up prompts

Continue the section

How would you encapsulate a model provider inside an application?

Practice focus

Answer path

Overview

Review signals

Follow-up prompts

Continue the section