Skip to content
tezvyn:

Deployment

130 bites tagged Deployment — interview questions with model answers, and 60-second explainers.

Python & FastAPI2 min read

Why use an ASGI server like Uvicorn instead of the dev server?

Tests: dev vs prod environment distinction. Answer: Uvicorn is a prod server program; contrast dev server's constant restart/break/fix cycle with prod needs for performance, stability, and uninterrupted access.

Python & FastAPI2 min read

FastAPI Container Build and Deploy Pipeline

Treat the Docker image as the immutable artifact: one build runs everywhere. Deploy FastAPI workers behind a load balancer, one process per container. The footgun is baking secrets into the image or running multiple processes; that breaks horizontal scaling.

MLOps & Infrastructure2 min read

Explain a model registry's purpose and what to store per version

Tests if you treat the registry as a governance bridge between training and production, not just storage. Strong answers cite versioned artifacts, lineage, metrics, dependencies, and approval gates. Red flag: calling it a file dump or experiment tracker.

MLOps & Infrastructure2 min read

Expose a trained model as a simple web service

Practical MLOps knowledge from model serialization to serving. Package the model into a standard format, containerize it, expose a REST endpoint behind a load balancer, and add monitoring. A bare Flask server without containers or health checks is a red flag.

MLOps & Infrastructure2 min read

Explain model quantization, its benefits, drawbacks, and validation approach

Tests precision trade-offs in production. Answer: define lowering weights from fp32 to int8/int4; cite memory and latency gains versus accuracy loss; validate with downstream benchmarks and shadow A/B. Red flag: treating as lossless or skipping task metrics.

MLOps & Infrastructure2 min read

Deploy a trained model as a containerized REST API

This tests practical MLOps fluency. A strong answer covers loading the artifact, wrapping it in a web server, building a Dockerfile, and exposing health and predict endpoints. Red flag: conflating training with serving or omitting resource limits.

MLOps & Infrastructure2 min read

What is a model registry's purpose in CI/CD4ML and its CI/CD interaction?

Tests if you see the model registry as the bridge between experiments and production, not just storage. A strong answer explains how CI publishes validated artifacts and CD consumes versioned models. Red flag: calling it a passive file dump without versioning.

MLOps & Infrastructure2 min read

What automated tests belong in CI before deploying a classification model?

Name data schema checks, performance regression vs baseline, bias audits, and artifact integrity. Distinguishing code tests from ML-specific CI validation. Only testing the inference API while ignoring model behavior.

MLOps & Infrastructure2 min read

Design a CI/CD pipeline that automates model promotion from Staging to Production

Tests whether you treat model promotion as a gated software delivery workflow. Strong answers use registry state-change triggers, automated drift and performance checks, canary deployment gates, and rollback.

MLOps & Infrastructure2 min read

What are the primary differences between traditional DevOps and MLOps?

Tests if you know MLOps extends CI/CD to data and probabilistic assets. Contrast code artifacts with models and datasets. Distinguish commit triggers from drift or retraining triggers. Red flag: treating models as static binaries and ignoring data lineage.

MLOps & Infrastructure2 min read

MLflow Models Standardize Deployment Packaging

MLflow Models wrap artifacts into a standard package so one pipeline serves sklearn or PyTorch without new deployment code. Teams ship experiments to REST endpoints without Dockerfiles per model. Missing dependency logging lets model load but fail to predict.

MLOps & Infrastructure2 min read

Model Registry: Source of Truth for Deployed Models

A model registry is the source of truth for which trained model runs where, turning anonymous artifact files into versioned, staged assets. It matters when you deploy multiple models or need instant rollbacks.

MLOps & Infrastructure2 min read

Model Server: The MLOps Deployment Bridge

A model server bridges ML training and production, operationalizing models within your release cycle. Use it when models must become first-class CI/CD citizens. The footgun is treating deployment as a one-time handoff rather than repeatable infrastructure.

MLOps & Infrastructure2 min read

Shadow Deployment: Test Models on Real Traffic

Shadow deployment runs a new model on real traffic without serving its predictions, letting you catch data drift before users are affected. It is the safest production validation method, but teams often forget to monitor its latency and resource costs.

Data Science & Analytics2 min read

ML Model Registry: Source of Truth for Production Models

A model registry is version control for trained models, not just code. It tracks which artifact is running in production, who approved it, and how it was built. Skip it and you get untracked files in S3 with no way to reproduce a production model.

Data Science & Analytics2 min read

ML Pipeline: Systematic Model Delivery

A machine learning pipeline is the systematic workflow that carries models from data labeling through deployment inside MLOps. It keeps the AI lifecycle repeatable rather than ad hoc. The footgun is treating a one-off notebook as a production pipeline.

CI/CD & Automation2 min read

How do you add a basic post-deployment health check in CI/CD?

Tests deployment validation beyond exit-code success. Outline: add a post-deploy stage that probes an HTTP endpoint, checks status code and latency, validates critical dependencies, and triggers rollback on failure.

CI/CD & Automation2 min read

Walk me through deploying a new version using a GitOps workflow

Merge updates manifests; the GitOps controller compares desired state in Git to live state and reconciles. Separation of CI from CD and Git as the single source of truth. Mixing CI with CD or recommending manual deploy.

CI/CD & Automation2 min read

How would you use distributed tracing to debug a deployment latency issue?

This tests causal request-path analysis beyond aggregate metrics. A strong answer filters traces by the new version, finds the exact regressed span, and compares it to a pre-deployment baseline.

CI/CD & Automation2 min read

Design an automated rollback process when deployment error rates spike

Gate on error-rate and latency thresholds; use blue-green deploys to limit blast radius; require human approval for stateful rollbacks. judgment when automating rollbacks. instant full rollback from one noisy metric.

CI/CD & Automation2 min read

What are the four Golden Signals for service health monitoring?

Tests whether you can name the four essential metrics—latency, traffic, errors, saturation—and explain why each matters for detecting user-facing regressions after a release, rather than drowning in infrastructure noise.

CI/CD & Automation2 min read

Blue/green deployment fails during switch-over with partial decommissioning; recovery and process changes?

Tests whether you can recover when a blue/green rollback path is compromised. Strong answers stop the bleed, revive blue if possible, and mandate keeping blue fully warm until green is stable. Red flag: "just roll back" ignoring partial decommissioning.

CI/CD & Automation2 min read

What are the major challenges of blue/green deployments with relational databases?

Tests decoupling schema and code changes in stateful blue/green deployments. Strong answers cover the additive-then-deletive pattern, backward compatibility for both app versions, and shared DB risks. Red flag: split DBs with no rollback or sync plan.

CI/CD & Automation2 min read

What problem can a breaking API change cause during a rolling update?

Tests if you know rolling updates run mixed versions, so breaking API changes crash cross-traffic. Good answer: note old and new pods serve together, watch probes fail, and monitor 5xx spikes. Red flag: claiming Kubernetes isolates versions during rollout.

Get Deployment bites daily.

Five a day, five minutes, offline. With quizzes so it sticks.

Open testing — you’ll join as an early tester.