Loading Engineering Foundry…
Connect a product decision to data, labels, evaluation, training, serving, rollout, monitoring, and risk.
No recent ML practice in this browser yet. Start with framing or choose a role-focused path.
Basic System Design concepts help, but they do not gate access. Each concept links only the specific foundation it relies on.
Use one repeatable sequence, returning to an earlier decision whenever later evidence changes it.
The core path connects each decision to a concept lesson and related practice.
Name the user, action, prediction unit, constraints, baseline, and whether ML is justified.
Connect business and user outcomes to task metrics, slices, and guardrails.
Trace exposure and events into labels, point-in-time examples, representations, and offline evidence.
Separate offline training from the online decision path and make timing, state, cost, and fallback explicit.
Use offline gates, shadow, canary, experiments where appropriate, staged rollout, and rollback.
Monitor data, system, predictions, and outcomes; diagnose before retraining or expanding complexity.
This check stays in this tab and does not determine access. If two or more feel unfamiliar, review the linked System Design foundations as you go.
0 of 4 checked. This is orientation, not a readiness score.
Translate a product need into a measurable prediction or decision.
2 concepts · 2 connected problemsSeparate offline model quality, online impact, and system guardrails.
3 concepts · 2 connected problemsDesign labels and features without leaking future information.
5 concepts · 3 connected problemsStart with a baseline and add complexity only where evidence supports it.
4 concepts · 3 connected problemsMake experiments and retraining repeatable, observable, and recoverable.
3 concepts · 2 connected problemsMeet prediction freshness and latency needs safely in production.
4 concepts · 3 connected problemsDetect system degradation before business outcomes arrive.
4 concepts · 3 connected problemsApply the framework to retrieval, ranking, embeddings, and generative systems.
4 concepts · 6 connected problemsThese branches reorder the same core material. They do not skip formulation, evaluation, reliability, or risk.
MLE · SWE-ML · Applied Scientist
MLE · SWE-ML
Applied Scientist · MLE
SWE-ML
Open an Engineering Foundry dossier for a task-specific walkthrough and guided, untimed, or timed rehearsal. Optional videos and articles are further learning.
13 designs
Retrieve and rank useful items while balancing relevance, discovery, inventory health, safety, and feedback loops.
Rank rapidly changing content while balancing session relevance, recency, quality, diversity, abuse, and ecosystem effects.
Combine lexical and learned retrieval with query understanding, ranking, permissions, freshness, and honest fallback.
Serve safe, fresh prefix suggestions within an every-keystroke latency budget.
Retrieve eligible ads and combine calibrated predictions with pacing, value, auction policy, and user guardrails.
Score rare adversarial transactions with streaming features, calibrated action bands, review, appeals, and safe fallback.
Separate detection from policy, human review, enforcement, appeal, and evolving adversarial response.
Correct a strong routing baseline with fresh spatiotemporal signals while protecting tail error and fallback.
Produce multi-horizon forecast distributions with rolling validation and operational cost awareness.
Design a shared temporal feature platform only when reuse, point-in-time correctness, and online delivery earn it.
Coordinate reproducible training, evaluation, registry, promotion, quotas, and scheduled or event-driven operation.
Serve a heterogeneous model fleet with explicit SLOs, routing, batching, autoscaling, compatibility, admission, and graceful degradation.
Join permission-aware retrieval and generation with separate evaluation, citations, injection controls, cost, and fallback.
Every lesson includes a scenario, mechanism, alternatives, consequences, failure modes, worked example, exercise, level overlays, System Design prerequisites, and reviewed sources.
Open the concept directory