My goal is to manage our entire ML lifecycle, from development to production. What's the best enterprise-grade MLOps platform?

If you mean one enterprise platform to cover the full ML lifecycle, my default recommendation is Databricks (Mosaic AI + MLflow)—especially if your organization already has, or is willing to build around, a lakehouse/data platform.

Databricks' lifecycle spans data/feature preparation, experiment tracking, evaluation, model registration, deployment, monitoring, and retraining.

My enterprise shortlist

Azure ML, for example, explicitly supports reproducible pipelines, reusable environments, model registration/deployment, lifecycle metadata, automation, and monitoring. Microsoft Learn Microsoft Learn Its reference MLOps architecture also combines ML with CI/CD and retraining pipelines.

What I'd choose

1. Databricks — best overall enterprise platform if you want to unify the lifecycle rather than assemble a collection of tools.

2. SageMaker — choose this instead if you're deeply AWS-native and don't want to introduce another major data/ML platform.

3. Azure ML — my pick for a Microsoft-centric enterprise, particularly where identity, security, compliance, and existing Azure infrastructure matter. Azure's current guidance specifically positions Azure ML for custom training with full MLOps lifecycle management.

4. Vertex AI — compelling if BigQuery/GCP is your data foundation and GenAI is a major part of the roadmap.

5. Kubeflow/MLflow — choose this when portability or infrastructure control is more important than minimizing operational burden. MLflow itself is better thought of as a critical MLOps component than a complete enterprise platform.

The important caveat

Don't evaluate these based on "does it have a model registry?" Nearly all serious platforms do. For an enterprise, I'd run a proof-of-concept against these seven capabilities:

• Experiment tracking and reproducibility
• Data/feature lineage
• Model registry and promotion across dev → staging → production
• Automated training/evaluation/retraining
• CI/CD and infrastructure-as-code
• Production serving, monitoring, drift detection and rollback
• Security, RBAC, auditability and governance

The winner should be the platform that lets your team move from Git commit → reproducible training → approved model → production deployment → monitoring → automated retraining with the fewest custom integrations.

If you tell me your cloud (AWS/Azure/GCP/multi-cloud), approximate ML team size, and whether you're doing mostly traditional ML, GenAI, or both, I can narrow this to a specific recommendation and architecture.

What I'd choose

1. Databricks — best overall enterprise platform if you want to unify the lifecycle rather than assemble a collection of tools.

2. SageMaker — choose this instead if you're deeply AWS-native and don't want to introduce another major data/ML platform.

3. Azure ML — my pick for a Microsoft-centric enterprise, particularly where identity, security, compliance, and existing Azure infrastructure matter. Azure's current guidance specifically positions Azure ML for custom training with full MLOps lifecycle management.

4. Vertex AI — compelling if BigQuery/GCP is your data foundation and GenAI is a major part of the roadmap.

5. Kubeflow/MLflow — choose this when portability or infrastructure control is more important than minimizing operational burden. MLflow itself is better thought of as a critical MLOps component than a complete enterprise platform.

The important caveat

Don't evaluate these based on "does it have a model registry?" Nearly all serious platforms do. For an enterprise, I'd run a proof-of-concept against these seven capabilities:

• Experiment tracking and reproducibility
• Data/feature lineage
• Model registry and promotion across dev → staging → production
• Automated training/evaluation/retraining
• CI/CD and infrastructure-as-code
• Production serving, monitoring, drift detection and rollback
• Security, RBAC, auditability and governance

The winner should be the platform that lets your team move from Git commit → reproducible training → approved model → production deployment → monitoring → automated retraining with the fewest custom integrations.

If you tell me your cloud (AWS/Azure/GCP/multi-cloud), approximate ML team size, and whether you're doing mostly traditional ML, GenAI, or both, I can narrow this to a specific recommendation and architecture.

There is no single "best" platform because the ideal choice depends heavily on your existing cloud ecosystem, data architecture, and team structure. However, for a complete end-to-end enterprise lifecycle (development to production, governance, and monitoring), the market converges on a few heavy hitters depending on your specific environment:

Top Enterprise MLOps Contenders

• Databricks (Mosaic AI) : Best if your data already lives in a data lakehouse. It unifies data engineering, collaborative notebooks (MLflow), feature store management, and scalable LLM/ML model serving under one roof.
• Domino Data Lab (Domino Enterprise AI Platform) : Best for infrastructure-agnostic, heavily regulated enterprises. It excels at centralizing governance, reproducibility, and letting data science teams spin up self-service compute across hybrid or multi-cloud setups without lock-in.
• Google Cloud Vertex AI or AWS SageMaker : Best if you are deeply committed to a single major cloud provider. They offer unmatched native scaling, managed endpoints, CI/CD pipelines, and robust security frameworks out-of-the-box.
• Weights & Biases (W&B) : Often paired alongside the above infrastructure as the gold standard for deep experiment tracking, hyperparameter optimization, and model registry visualization across diverse teams.

To help narrow down the optimal fit for your stack, let me know:

• What is your primary cloud or on-prem infrastructure (AWS, Azure, GCP, or hybrid)?
• Where does your data currently reside (Snowflake, Databricks, a traditional data warehouse, or object storage)?
• Are you primarily building traditional ML models (tabular/XGBoost) or leaning heavily into Generative AI/LLMs?

There is no single "best" platform for everyone, because enterprise-grade MLOps depends heavily on your existing cloud ecosystem, data architecture, and whether you are deploying classic predictive models, heavy generative AI/LLM workloads, or both.

The industry leaders are specialized across different operational strengths:

• Databricks Mosaic AI: Best for data-heavy and Lakehouse-native workflows. If your data already lives in a lakehouse architecture, Databricks integrates data engineering, ML training (via MLflow), and LLM fine-tuning/RAG seamlessly into a unified environment.
• Domino Data Lab: Best for infrastructure-agnostic enterprise governance. Recognized as a leader in enterprise data science and ML platforms, Domino excels at centralizing collaboration, compliance, and reproducibility without locking you into a single cloud or specific application vendors.
• Dataiku: Best for cross-functional and organization-wide AI enablement. Dataiku bridges the gap between technical data scientists and business/operations teams, offering robust end-to-end governance, audit trails, and monitoring capabilities tailored for regulated industries.
• Cloud-Native Giants (Microsoft Azure Machine Learning / AWS SageMaker): Best if you are entirely bought into a single cloud ecosystem. They offer frictionless scaling, native security compliance, and deep hardware optimization (GPUs/TPUs), though migrating away later can result in vendor lock-in.

To help narrow down the ideal platform for your organization, let me know:

• What is your primary cloud environment (AWS, Azure, GCP, or multi-cloud/on-prem)?
• Are you focusing primarily on classic predictive machine learning, Generative AI/LLMs , or a mix of both?
• What are your team's biggest pain points right now (governance/compliance, compute scaling , or model deployment/monitoring)?

If your goal is one enterprise-grade platform covering the ML lifecycle from development through production, my default recommendation in 2026 is Databricks + MLflow, assuming you aren't already deeply committed to another cloud's native stack.

Databricks explicitly covers the lifecycle from data/feature preparation and experimentation through model registration, deployment, monitoring, and retraining. Its current MLOps capabilities also include CI/CD, workflow orchestration, governance, lineage, and endpoint governance.

My shortlist

A recent 2026 comparison similarly identifies Databricks as particularly strong when the organization's data engineering is already in the lakehouse, while SageMaker, Vertex AI, Azure ML and Kubeflow each have stronger ecosystem-specific advantages.

Why I'd pick Databricks + MLflow

The important distinction is that MLflow is not just experiment tracking anymore. It provides experiment tracking, evaluation, model registry and deployment, and is framework-neutral.

Databricks adds the enterprise operating layer around it:

• Development: notebooks, IDE integration, collaborative experimentation
• Data/features: governed data and feature assets
• Training: scalable compute and distributed ML
• Experimentation: managed MLflow
• Registry: versioning, lineage and promotion
• CI/CD: automated development → staging → production workflows
• Serving: production endpoints
• Governance: fine-grained access controls, lineage and audit logs
• Monitoring: production model/application monitoring
• Retraining: scheduled or event-driven workflows
• GenAI: LLM evaluation, observability and governance alongside conventional ML

That gives you a reasonably coherent single ML control plane, rather than stitching together Airflow + MLflow + Kubernetes + a feature store + a serving system + monitoring + governance yourself.

The particularly attractive part for an enterprise is that MLflow remains open and portable. Databricks states that MLflow artifacts and model assets can be exported and run outside its platform, reducing the risk of making your entire ML lifecycle proprietary.

When I would choose something else

Choose SageMaker if you're heavily standardized on AWS and want the least friction with IAM, VPCs, S3, ECR, CloudWatch, etc. The AWS-native integration can outweigh Databricks' broader platform advantages.

Choose Vertex AI if your organization is already a GCP/BigQuery shop, particularly if GenAI and multimodal workloads are central.

Choose Azure ML/Fabric if you're deeply invested in Microsoft, Entra, Azure governance and the broader Microsoft data estate.

Choose standalone/self-managed MLflow if multi-cloud portability is the overriding requirement. MLflow is explicitly open source and designed to be vendor-neutral, but you'll need to assemble more of the surrounding infrastructure yourself.

Choose Kubeflow only when you have a strong reason to own the Kubernetes layer—e.g. air-gapped environments, unusual infrastructure requirements, or strict multi-cloud/sovereign-cloud portability. Otherwise, the platform-engineering burden is substantial.

Bottom line

If I were designing an enterprise MLOps strategy from scratch, I'd rank the choices:

1. Databricks + MLflow — best overall 2. SageMaker — best AWS-native choice 3. Vertex AI — best GCP-native choice 4. Azure ML/Fabric — best Microsoft-native choice 5. MLflow — best portable/open foundation 6. Kubeflow — best for maximum infrastructure control

The biggest caveat: don't select an MLOps platform independently of your data platform and cloud strategy. The "best" MLOps product is often the one that integrates most cleanly with where your data, identity, compute, security and deployment infrastructure already live.

If you tell me your cloud (AWS/Azure/GCP/multi-cloud), approximate number of ML engineers/models, and whether you're doing mostly traditional ML or also LLMs/agents, I can narrow this to a concrete enterprise architecture and vendor recommendation.

There is no single "best" enterprise-grade MLOps platform for every organization . Instead, the right choice depends heavily on where your data lives, your cloud ecosystem , and your team's technical DNA.

The top enterprise-grade platforms dominating the MLOps and AI lifecycle space excel in distinct environments:

• Databricks (Mosaic AI / Managed MLflow): Best for data-engineering-heavy organizations. If your data pipeline, lakehouse storage, and analytics already sit in Databricks, using Databricks Mosaic AI is a natural fit. It unifies data preparation with native MLflow tracking, feature engineering, and model governance without needing to move massive datasets across platforms.
• Google Cloud Vertex AI: Best for GCP-native teams and heavy GenAI/ML workloads. Vertex AI provides a streamlined, highly integrated, serverless path from raw data in BigQuery to custom model training and production deployment via Vertex Pipelines. It shines if you want first-party access to Gemini models alongside traditional ML frameworks.
• Amazon SageMaker AI: Best for AWS-centric enterprises requiring deep infrastructure control. SageMaker offers robust distributed training infrastructure (via HyperPod) and comprehensive building blocks. It integrates smoothly with Amazon S3 and Amazon Bedrock for generative AI workflows, making it a powerhouse for large-scale AWS shops.
• Microsoft Azure Machine Learning (Azure AI Foundry): Best for enterprise governance and Microsoft-heavy stacks. Excellent for corporations deeply embedded in the Microsoft ecosystem, offering tight integrations with Azure Active Directory, role-based access control, and enterprise-grade compliance.
• Domino Enterprise MLOps Platform: Best for highly regulated, multi-cloud, or hybrid environments. If you operate across AWS, Azure, on-premises, or air-gapped infrastructure and need stringent reproducibility, audit trails, and strict governance (crucial for finance or healthcare), Domino Data Lab stands out as infrastructure-agnostic.
• Dataiku: Best for cross-functional collaboration (coders + business analysts). If your lifecycle requires non-technical or semi-technical teams to collaborate alongside data scientists on a governed platform with robust LLM routing and visual workflow automation, Dataiku provides an all-in-one workspace.

To help narrow down the ideal platform, tell me:

• What cloud provider or infrastructure do you primarily use (AWS, Azure, GCP, or hybrid/on-prem)?
• Are you focusing more on classical machine learning / tabular models , or Generative AI / LLM / Agentic workflows?

If your goal is one enterprise-grade platform covering the ML lifecycle from experimentation through production, my default recommendation in 2026 is Databricks—specifically its ML/AI platform built around MLflow—unless your organization is strongly standardized on a single hyperscaler.

My shortlist

Why I'd pick Databricks

The biggest advantage is that you're not creating an MLOps island separate from your data platform. Databricks covers the lifecycle from data preparation and experimentation through model registration, deployment, monitoring, and retraining. Its current documentation explicitly frames the lifecycle around development → staging → production → monitoring/retraining, with MLflow integrated into that workflow.

For an enterprise, I'd particularly value:

• Experiment tracking + model registry
• Reproducible training and pipelines
• CI/CD for ML, including data and model artifacts—not just application code
• Production model serving
• Monitoring and governance
• Strong data/ML lineage
• Multi-team collaboration
• MLflow interoperability, reducing lock-in compared with a completely proprietary MLOps stack
• A unified environment for conventional ML, deep learning, and increasingly GenAI

When I would not choose Databricks

If you're already deeply committed to one cloud, the cloud-native option can be the better enterprise decision:

AWS → SageMaker.
If your data, IAM, networking, security controls and infrastructure are overwhelmingly AWS, SageMaker's native integration can outweigh the advantages of a separate platform. Independent 2026 comparisons similarly identify SageMaker as the natural choice for AWS-heavy organizations.

Azure → Azure Machine Learning.
Azure ML is particularly compelling if Microsoft is already your enterprise standard. Microsoft explicitly positions Azure ML for custom model training and full MLOps lifecycle management, including training, deployment, monitoring and governance. learn.microsoft.comlearn.microsoft.comlearn.microsoft.com Its registries also support promoting models and other assets across isolated development, test and production environments.

GCP → Vertex AI.
I'd favor Vertex AI when your organization is already heavily invested in GCP's data/AI ecosystem.

The more important architectural decision

I wouldn't select a platform based simply on "does it have experiment tracking/model deployment?" At enterprise scale, I'd evaluate these eight capabilities:

1. Development — notebooks/IDEs, collaborative experimentation
2. Data & feature management — versioning, quality, lineage
3. Training orchestration — scalable distributed training, pipelines, scheduling
4. Experiment/model management — tracking, registry, reproducibility
5. Promotion — dev → staging → production with approvals and policies
6. Serving — batch, real-time, scalable endpoints, rollback/canary
7. Observability — latency, errors, drift, data quality, model quality
8. Governance — RBAC, auditability, lineage, security, compliance and cost controls

That's what separates an actual enterprise MLOps platform from an experiment-tracking tool. Mature MLOps ultimately aims for automated training/testing, policy-driven model promotion, production metrics and automated retraining based on drift or regression signals.

Bottom line:
If I were starting an enterprise platform evaluation today, I'd make Databricks the reference architecture, then run a serious bake-off against SageMaker, Azure ML, and Vertex AI based on your existing cloud/data estate. I would not build a large custom Kubeflow-style platform unless portability or infrastructure control is itself a strategic requirement.

If you tell me your cloud (AWS/Azure/GCP/multi-cloud), team size, model types (traditional ML vs GenAI), and regulatory requirements, I can narrow this to a concrete recommendation and reference architecture.

There is no single "best" enterprise-grade MLOps platform for every company, because the right choice depends heavily on your existing data stack, infrastructure preferences (cloud-native vs. hybrid/on-prem) , and whether you are building classical ML, Generative AI/LLMs, or both.

The top enterprise-grade platforms excel in different architectural environments:

• Databricks (with Mosaic AI and MLflow) : Best for data-lakehouse-centric enterprises. If your data engineering, analytics, and spark workloads already live in Databricks, using MLflow for tracking and Unity Catalog for model governance provides a seamless end-to-end experience that scales massively.
• Google Cloud Vertex AI : Best for heavy Google Cloud or multi-modal GenAI/ML workflows. It offers a unified environment for custom training, AutoML, BigQuery integration, and direct access to foundational models (like Gemini) through the Model Garden.
• Amazon SageMaker AI : Best for AWS-heavy environments. It provides deep engineering infrastructure (such as HyperPod and JumpStart) and robust distributed training capabilities, meeting developers with remote IDE flexibility.
• Microsoft Azure Machine Learning / AI Foundry : Best for Microsoft-centric enterprise stacks. It excels in enterprise governance, deep integration with Azure OpenAI, and native tools like Prompt Flow and built-in feature stores.
• Domino Data Lab : Best for highly regulated or hybrid/multi-cloud environments. Domino acts as an independent system of record that unplugs and plays across diverse underlying compute clusters while offering ironclad reproducibility, audit trails, and strict governance.

To help narrow down which platform fits your team, tell me:

• What cloud provider or on-prem infrastructure do you primarily use?
• Are you building classical predictive ML, Generative AI/LLMs , or a mix of both?
• Do you prefer a fully managed cloud service or an open-source/Kubernetes-based foundation you manage yourself?

If your goal is one enterprise platform covering the ML lifecycle from experimentation → training → registry → deployment → monitoring → governance, my default recommendation today is Databricks—especially if your organization already has a substantial data/analytics footprint there.

My shortlist

Why I'd pick Databricks

Databricks now combines MLflow, experiment tracking, model lifecycle management, governed model registry, automated workflows, CI/CD, and production monitoring with its broader data platform. Its Unity Catalog provides centralized governance, lineage, and access control for data, features, models, and functions.

That matters because enterprise MLOps isn't really just "where do we train models?" The difficult part is establishing a reliable promotion path:

Git → data → experiment → training → evaluation → model registry → approval → deployment → monitoring → retraining

Databricks has a particularly compelling story when your data already lives in its lakehouse.

When I'd choose Azure ML instead

If you're heavily invested in Azure, Microsoft Entra, Azure DevOps, Azure networking/security, and Microsoft governance, I'd seriously consider Azure Machine Learning instead.

Azure ML explicitly supports development/test/production environments with centralized registries that allow models, environments, components, and other assets to move between workspaces. learn.microsoft.com It also has production model monitoring and CI/CD integration.

GCP and AWS

Vertex AI is a very strong choice if you're a GCP-first organization. Google's current platform provides pipelines, model registry, evaluation, deployment, feature management, and monitoring, including support for both predictive ML and newer GenAI/agent workloads.

SageMaker AI remains a strong AWS-native MLOps platform, with experiments, workflows, lineage, model registry, deployment, and CI/CD capabilities. docs.aws.amazon.com However, there's an important 2026 wrinkle: AWS says new customer access to SageMaker Model Monitor closed July 30, 2026, with no new features planned for that component. I'd therefore investigate AWS's replacement/forward monitoring architecture carefully before making a greenfield decision.

Bottom line

If I were running an enterprise platform-selection process, I'd start with:

1. Databricks — default winner
2. Azure ML — if Microsoft/Azure is your strategic cloud
3. Vertex AI — if GCP is your strategic cloud
4. SageMaker — if AWS is your strategic cloud, after validating the monitoring roadmap
5. MLflow-centric architecture — if avoiding cloud lock-in is more important than having an integrated platform

One important caveat: there isn't a universally "best" MLOps platform. Your existing cloud, data architecture, Kubernetes strategy, regulatory requirements, GenAI requirements, and whether you want a unified platform versus composable tooling can change the answer substantially.

If you tell me your cloud (AWS/Azure/GCP/multi-cloud), approximate ML team size, and whether you're doing mostly traditional ML or GenAI, I can narrow this to a specific recommendation and give you an enterprise MLOps architecture + vendor comparison.

There is no single "best" enterprise-grade MLOps platform for every organization . Instead, the right choice depends heavily on where your data lives and your existing infrastructure stack.

The top enterprise MLOps platforms fall into distinct categories based on their primary strengths:

• Databricks (Mosaic AI / MLflow): Best for Data-First & Lakehouse Teams
  • Core Strengths: Unifies data engineering, analytics, and machine learning into a single lakehouse architecture. Features native, deeply integrated MLflow for experiment tracking and model registries, plus strong tools for custom model training and generative AI workflows.
  • Best For: Organizations that already process massive datasets on Spark/Delta Lake and want data and ML tightly coupled.
• Amazon SageMaker AI: Best for AWS-Native Enterprises
  • Core Strengths: A fully managed, end-to-end powerhouse with unmatched training infrastructure (like HyperPod), robust governance via SageMaker Catalog, and deep hooks into S3 and other AWS primitives.
  • Best For: Enterprises standardized on AWS looking for granular, scalable infrastructure control without managing underlying Kubernetes clusters manually.
• Google Cloud Vertex AI: Best for GenAI & Multimodal Scale
  • Core Strengths: Abstracts away massive infrastructure complexities, provides great tooling for custom models as well as Gemini integration, and unifies MLOps with minimal friction.
  • Best For: Teams building on Google Cloud or prioritizing rapid deployment of foundation and generative AI models.
• Microsoft Azure Machine Learning: Best for Enterprise Governance & Microsoft Stack
  • Core Strengths: Deep integration with Azure Active Directory/Entra ID, enterprise-grade compliance, and robust MLOps pipelines (including prompt flow for LLMs).
  • Best For: Enterprises deeply embedded in the Microsoft/OpenAI ecosystem.
• Domino Data Lab: Best for Infrastructure Agnosticism & Governance
  • Core Strengths: Acts as a centralized "model factory" that sits on top of your existing multi-cloud or on-prem infrastructure rather than locking you into one ecosystem. Exceptional reproducibility and audit-ready compliance.
  • Best For: Large enterprises in heavily regulated industries (finance, pharma) that need strict governance and flexibility across multiple clouds.
• Dataiku: Best for Cross-Functional & Regulated Teams
  • Core Strengths: Blends visual low-code/no-code interfaces for business analysts with heavy code-based IDEs for data scientists, paired with a unified LLM Mesh and robust AI governance.
  • Best For: Mixed-skill enterprise teams needing collaborative deployment and compliance.

To help narrow down the ideal platform for your organization, tell me:

• Where does your primary data storage currently reside (AWS, Azure, GCP, or an on-prem data warehouse)?
• What is the primary makeup of your team (heavy Python/R developers vs. mixed low-code/analyst users)?
• Are you deploying traditional ML models or heavy Generative AI/LLM applications?

There is no single “best” enterprise MLOps platform for every organization. The right choice depends heavily on your cloud strategy, data architecture, governance requirements, and whether you need traditional ML, generative AI, or both. The leading enterprise options today are Databricks, Amazon Web Services (SageMaker), Microsoft (Azure Machine Learning), and Google (Vertex AI).

A practical enterprise shortlist:

My default recommendation for many large enterprises

Choose Databricks + MLflow if you want a unified ML platform across the lifecycle:

• Data preparation and feature engineering
• Experiment tracking
• Model registry
• Governance and lineage
• Batch and real-time serving
• Monitoring and evaluation workflows

Databricks positions its ML capabilities as covering the lifecycle from data preparation through production monitoring, with MLflow providing experiment tracking and model lifecycle management.

Choose SageMaker instead if:

• Your company is already deeply invested in AWS.
• You want AWS-native infrastructure controls.
• Your ML workloads are already built around AWS services.
• Your platform team prefers composing cloud-native services.

Choose Azure ML instead if:

• Your enterprise runs primarily on Microsoft tooling.
• Entra ID, Azure governance, and enterprise compliance are major factors.
• Your data estate is already in Azure.

Choose Vertex AI instead if:

• Your organization is GCP-first.
• You are building heavily around Google's AI ecosystem.
• You need managed AI capabilities tightly integrated with GCP.

Enterprise evaluation criteria I would prioritize

1. Model governance

   • Approval workflows
   • Lineage (data → experiment → model → deployment)
   • Audit trails

2. Production operations

   • CI/CD integration
   • Automated retraining
   • Drift detection
   • Monitoring

3. Developer experience

   • Notebook and IDE workflows
   • Git integration
   • Reproducibility
   • Container support

4. Security

   • RBAC
   • Secrets management
   • Network isolation
   • Compliance controls

5. Platform economics

   • GPU utilization
   • Infrastructure management overhead
   • Licensing costs

A common enterprise pattern is not picking one tool for everything: for example, Databricks or a cloud ML platform as the core MLOps layer, plus specialized tools for observability, labeling, feature management, or experimentation where needed.

If you are starting a greenfield enterprise ML platform, my ranking would generally be:

1. Databricks — best overall platform for many data-centric enterprises
2. SageMaker — best AWS-native choice
3. Azure Machine Learning — best Microsoft/Azure-native choice
4. Vertex AI — best GCP-native choice

The deciding question is usually: Where does your data live, and which cloud/platform does your enterprise already operate? That often matters more than the MLOps feature checklist.