Design an AI-Powered Code Assistant — System Design Interview Practice

Interview prompt

Design an IDE and CI-integrated code assistant that serves low-latency completions, code generation, explanations, and review while protecting source code and respecting repository context.

• Define editor context boundaries, repository indexing, language parsing, prompt construction, completion cancellation, model versions, and telemetry consent.
• Keep inline completion bounded and fast while supporting multi-file retrieval, workspace isolation, secret/PII redaction, and context freshness.
• Separate interactive inference from indexing, batch code review, evaluation, and CI workflows; make requests cancellable and retries safe.
• Explain hallucination/security risks, licensing, prompt injection, data retention, rollout, observability, and a local/cache fallback.

Requirements and scale assumptions

• Provide inline and chat completions, natural-language code generation, explanations, refactoring, test generation, review comments, and CI annotations.
• Index authorized repositories with branch/version awareness, retrieve relevant context, stream/cancel responses, and show model/context metadata.
• Support tenant policies, repository deletion, telemetry opt-out, rate limits, model rollback, prompt-injection defenses, and dependency failure.
• Meet p95 first-token inline completion latency under 200ms for warm contexts and cap context retrieval/inference work.
• Scale to 1M active developers and large repository indexes without a single hot key or unbounded synchronous work.
• Do not lose committed state; make retries and duplicate events safe.
• Degrade safely when downstream workers, caches, or external dependencies fail.
• 1M developers, 100k concurrent IDE sessions, and million-file repositories
• Partition by the primary tenant, user, item, or geographic key and isolate hot partitions.
• Keep serving state bounded; retain raw events or durable records for replay and auditing.

• Peak scale: 1M developers; 100k IDE sessions — Capacity assumption that drives partitioning and backpressure.
• Latency target: inline p95 first token < 200ms — User-facing budget for the primary request or read path.
• Durable boundary: Committed before async — Authorized repository snapshots and model artifacts are authoritative; indexes and completions are derived.
• Async boundary: At-least-once workers — Keep Use code-specific LLMs (CodeLlama, StarCoder), Implement Fill-in-the-Middle (FIM) for completions, Use tree-sitter for language-aware context extraction off the synchronous path.

Key entities

• DatasetVersiondatasetId, version, schemaHash, qualityStatus, lineage, createdAt

  Immutable ai powered code assistant input version used for reproducible training, evaluation, or replay.

• FeatureSnapshotentityId, featureSetVersion, eventTime, values, sourceWatermarks

  Point-in-time ai powered code assistant features with source watermarks so online and offline values can be compared.

• TrainingRunrunId, datasetVersion, codeVersion, metrics, artifactUri, status

  Audited ai powered code assistant run that records data, code, dependency, and evaluation lineage.

• ModelVersionmodelId, version, stage, schema, qualityGates, endpoint

  A promotable ai powered code assistant model version with rollout state, contract, and rollback metadata.

Data flow

1. 1. Register and validate training dataThe ai powered code assistant gateway records an immutable dataset version, schema, lineage, quality status, and privacy disposition.
2. 2. Build point-in-time featuresFeature workers join ai powered code assistant inputs using event-time watermarks, prevent leakage, and publish the same feature contract for training and serving.
3. 3. Train and evaluate asynchronouslyThe orchestrator schedules ai powered code assistant runs with checkpointed artifacts, reproducible environments, and metrics tied to the exact input versions.
4. 4. Gate and serve a model versionA registry compares ai powered code assistant quality, bias, safety, and compatibility gates before canary or production rollout with an immediate rollback pointer.
5. 5. Monitor drift and learn from feedbackOnline inference records latency, errors, drift, and delayed labels so ai powered code assistant retraining is evidence-driven rather than triggered by guesswork.

Deep dives and trade-offs

• Reproducibility and leakage preventionPin ai powered code assistant data, feature, code, dependency, and model versions for every run. Use point-in-time joins and quarantine failed quality or privacy checks before training. Keep raw inputs and artifacts immutable so a result can be replayed after a dependency changes.
• Safe promotion and serving contractsSeparate ai powered code assistant model registration from deployment and require signed artifacts plus schema compatibility. Use shadow traffic, canaries, rollback pointers, and per-version latency/error budgets. Return model version and feature freshness so clients can explain or reproduce a prediction.
• Drift, feedback, and costMeasure feature drift, prediction drift, label delay, and segment-level quality for ai powered code assistant rather than only aggregate accuracy. Sample expensive inference and cap retraining concurrency with an explicit GPU or compute budget. Keep human corrections and delayed labels linked to the original prediction and model version.
• Batch versus online featuresPrefer a shared feature contract with batch backfills and a low-latency online serving path for decisions that need freshness. Two independently defined transformations create training-serving skew and hard-to-debug regressions.
• Synchronous versus asynchronous inferenceKeep interactive ai powered code assistant inference synchronous within a strict budget and queue large or expensive jobs. A request path that waits for model loading, enrichment, or retraining turns downstream slowness into an outage.
• Global model versus segment modelsStart with one versioned model and add segment-specific models only when quality or policy evidence justifies the operational cost. Many simultaneously active versions multiply monitoring, rollback, and data-lineage burden.