Public task directory

Technical interview practice tasks

Every entry below is a crawlable link to a unique public task page. This directory complements the interactive filters and keeps the complete catalog connected through standard pagination.

Page 186 of 386 · 18,491 tasks

ML System Design · Hard · 55 min

Design point-in-time definitions, streaming and batch paths, freshness SLOs, backfills, access controls, online serving, and skew monitoring

Design point-in-time definitions, streaming and batch paths, freshness SLOs, backfills, access controls, online serving, and skew monitoring.

View task

ML System Design · Hard · 55 min

Design candidate generation, ranking, feature freshness, feedback, exploration, serving fallbacks, and delayed-outcome monitoring for 60k requests per second

Design candidate generation, ranking, feature freshness, feedback, exploration, serving fallbacks, and delayed-outcome monitoring for 60k requests per second.

View task

ML System Design · Hard · 55 min

Design reproducible backfills, hierarchical forecasts, feature lineage, scheduled serving, overrides, monitoring, and model promotion across thousands of stores

Design reproducible backfills, hierarchical forecasts, feature lineage, scheduled serving, overrides, monitoring, and model promotion across thousands of stores.

View task

ML System Design · Hard · 55 min

Design ingestion, retrieval, prompt and model versioning, citation checks, safety policy, evaluation, caching, fallback, and human feedback

Design ingestion, retrieval, prompt and model versioning, citation checks, safety policy, evaluation, caching, fallback, and human feedback.

View task

ML System Design · Hard · 55 min

Design routing, batching, autoscaling, versioning, canaries, fallbacks, tenancy, cost controls, and observability for GPU and CPU models

Design routing, batching, autoscaling, versioning, canaries, fallbacks, tenancy, cost controls, and observability for GPU and CPU models.

View task

ML System Design · Hard · 55 min

Design retrieval, reranking, online features, query understanding, cache and fallback behavior, evaluation, rollout, and tail-query monitoring

Design retrieval, reranking, online features, query understanding, cache and fallback behavior, evaluation, rollout, and tail-query monitoring.

View task

ML System Design · Hard · 55 min

Design multimodal ingestion, policy labels, human review, regional handling, threshold policy, serving degradation, appeals, and delayed-label monitoring

Design multimodal ingestion, policy labels, human review, regional handling, threshold policy, serving degradation, appeals, and delayed-label monitoring.

View task

ML Take Home · Hard · 90 min

For each entity keep the row with the greatest event_time not after cutoff; return rows sorted by entity

For each entity keep the row with the greatest eventtime not after cutoff; return rows sorted by entity.

View task

ML Take Home · Hard · 90 min

Return counts for passed, failed, and blocked rows. blocked takes precedence over passed

Return counts for passed, failed, and blocked rows. blocked takes precedence over passed.

View task

ML Take Home · Hard · 90 min

Return precision_at_k for each query as [relevant_in_top_k,k], preserving query order. Each query contains ranked ids and relevant ids

Return precisionatk for each query as [relevantintopk,k], preserving query order. Each query contains ranked ids and relevant ids.

View task

ML Take Home · Hard · 90 min

Return the minimum replica count so that replicas * qps_per_replica * target_utilization is at least peak_qps

Return the minimum replica count so that replicas qpsperreplica targetutilization is at least peakqps.

View task

Model Incident · Medium · 35 min

Home-feed engagement fell 18% within 25 minutes of model v184 reaching 60% traffic, while CPU, error rate, and serving latency remain healthy. A feature-schema change, a catalog backfill, and a ranking-policy rollout happened in the same

Home-feed engagement fell 18% within 25 minutes of model v184 reaching 60% traffic, while CPU, error rate, and serving latency remain healthy. A feature-schema change, a catalog backfill.

View task

ML Notebook Lab · Medium · 35 min

A subscription churn model reports 0.94 offline ROC-AUC but only 0.61 after deployment. The training extract contains 1.8M account snapshots, repeated customers, late-arriving support events, and fields that may only exist after

A subscription churn model reports 0.94 offline ROC-AUC but only 0.61 after deployment. The training extract contains 1.8M account snapshots, repeated customers, late-arriving support.

View task

ML System Design · Medium · 60 min

Design the production decision system for a global card issuer processing 18k authorization requests per second. The ML contribution has an 80 ms p99 budget, confirmed fraud labels arrive 7–45 days later, EU features and explanations must

Design the production decision system for a global card issuer processing 18k authorization requests per second. The ML contribution has an 80 ms p99 budget, confirmed fraud labels arrive.

View task

Coding (Platform) · Easy · 30 min

Project Directory Search

Workspace: Express Difficulty: Easy Time: 30 minutes Context The project directory already authenticates callers and returns a stable page of active projects. Its new search box sends a.

View task

Coding (Platform) · Hard · 90 min

Retry Safe Reservations

Workspace: Express Difficulty: Hard Time: 90 minutes Context A project reservation endpoint already handles a single successful request. Under duplicate delivery or concurrent requests, it.

View task

Coding (Platform) · Medium · 60 min

Tenant Project Pagination

Workspace: Express Difficulty: Medium Time: 60 minutes Context Two customer accounts use the same project service. An incident report shows another tenant’s projects in a search response and.

View task

Coding (Platform) · Easy · 30 min

Project Directory Search

Workspace: Fastify Difficulty: Easy Time: 30 minutes Context The project directory already authenticates callers and returns a stable page of active projects. Its new search box sends a.

View task

Coding (Platform) · Hard · 90 min

Retry Safe Reservations

Workspace: Fastify Difficulty: Hard Time: 90 minutes Context A project reservation endpoint already handles a single successful request. Under duplicate delivery or concurrent requests, it.

View task

Coding (Platform) · Medium · 60 min

Tenant Project Pagination

Workspace: Fastify Difficulty: Medium Time: 60 minutes Context Two customer accounts use the same project service. An incident report shows another tenant’s projects in a search response and.

View task

Coding (Platform) · Easy · 30 min

Project Directory Search

Workspace: NestJS Difficulty: Easy Time: 30 minutes Context The project directory already authenticates callers and returns a stable page of active projects. Its new search box sends a title.

View task

Coding (Platform) · Hard · 90 min

Retry Safe Reservations

Workspace: NestJS Difficulty: Hard Time: 90 minutes Context A project reservation endpoint already handles a single successful request. Under duplicate delivery or concurrent requests, it.

View task

Coding (Platform) · Medium · 60 min

Tenant Project Pagination

Workspace: NestJS Difficulty: Medium Time: 60 minutes Context Two customer accounts use the same project service. An incident report shows another tenant’s projects in a search response and.

View task