diff --git a/Commercial Support.md b/COMMERCIAL_SUPPORT.md
similarity index 100%
rename from Commercial Support.md
rename to COMMERCIAL_SUPPORT.md
diff --git a/README.md b/README.md
index d23b196e969..14ad32e0b33 100644
--- a/README.md
+++ b/README.md
@@ -22,48 +22,57 @@
## Overview
-Litmus is an open source Chaos Engineering platform that enables teams to identify weaknesses & potential outages in infrastructures by inducing chaos tests in a controlled way.
+LitmusChaos is an open source Chaos Engineering platform that enables teams to identify weaknesses & potential outages in infrastructures by
+inducing chaos tests in a controlled way. Developers & SREs can practice Chaos Engineering with Litmus as it is easy to use, based on modern
+chaos engineering principles & community collaborated. It is 100% open source & a CNCF project.
-Developers & SREs can simply execute Chaos Engineering with Litmus as it is easy to use, based on modern chaos engineering practices & community collaborated.
-Litmus is 100% open source & CNCF-hosted.
+Litmus takes a cloud-native approach to create, manage and monitor chaos. The platform itself runs as a set of microservices and uses Kubernetes
+custom resources to define the chaos intent, as well as the steady state hypothesis.
-Litmus takes a cloud-native approach to create, manage and monitor chaos. Chaos is orchestrated using the following Kubernetes Custom Resource Definitions (**CRDs**):
+At a high-level, Litmus comprises of:
-- **ChaosEngine**: A resource to link a Kubernetes application or Kubernetes node to a ChaosExperiment. ChaosEngine is watched by Litmus' Chaos-Operator which then invokes Chaos-Experiments
-- **ChaosExperiment**: A resource to group the configuration parameters of a chaos experiment. ChaosExperiment CRs are created by the operator when experiments are invoked by ChaosEngine.
-- **ChaosResult**: A resource to hold the results of a chaos-experiment. The Chaos-exporter reads the results and exports the metrics into a configured Prometheus server.
+- **Chaos Control Plane**: A centralized chaos management tool called chaos-center, which helps construct, schedule and visualize Litmus chaos workflows
+- **Chaos Execution Plane Services**: Made up of a chaos agent and multiple operators that execute & monitor the experiment within a defined
+ target Kubernetes environment.
-Chaos experiments are hosted on hub.litmuschaos.io. It is a central hub where the application developers or vendors share their chaos experiments so that their users can use them to increase the resilience of the applications in production.
+
-
+At the heart of the platform are the following chaos custom resources:
+
+- **ChaosExperiment**: A resource to group the configuration parameters of a particular fault. ChaosExperiment CRs are essentially installable templates
+ that describe the library carrying out the fault, indicate permissions needed to run it & the defaults it will operate with. Through the ChaosExperiment, Litmus supports BYOC (bring-your-own-chaos) that helps integrate (optional) any third-party tooling to perform the fault injection.
+
+- **ChaosEngine**: A resource to link a Kubernetes application workload/service, node or an infra component to a fault described by the ChaosExperiment.
+ It also provides options to tune the run properties and specify the steady state validation constraints using 'probes'. ChaosEngine is watched by the
+ Chaos-Operator, which reconciles it (triggers experiment execution) via runners.
+
+The ChaosExperiment & ChaosEngine CRs are embedded within a Workflow object that can string together one or more experiments in a desired order.
+
+- **ChaosResult**: A resource to hold the results of the experiment run. It provides details of the success of each validation constraint,
+ the revert/rollback status of the fault as well as a verdict. The Chaos-exporter reads the results and exposes information as prometheus metrics.
+ ChaosResults are especially useful during automated runs.
+
+ChaosExperiment CRs are hosted on hub.litmuschaos.io. It is a central hub where the
+application developers or vendors share their chaos experiments so that their users can use them to increase the resilience of the applications
+in production.
+
+
## Use cases
- **For Developers**: To run chaos experiments during application development as an extension of unit testing or integration testing.
-- **For CI pipeline builders**: To run chaos as a pipeline stage to find bugs when the application is subjected to fail paths in a pipeline.
-- **For SREs**: To plan and schedule chaos experiments into the application and/or surrounding infrastructure. This practice identifies the weaknesses in the system and increases resilience.
+- **For CI/CD pipeline builders**: To run chaos as a pipeline stage to find bugs when the application is subjected to fail paths in a pipeline.
+- **For SREs**: To plan and schedule chaos experiments into the application and/or surrounding infrastructure. This practice identifies the weaknesses
+ in the deployment system and increases resilience.
## Getting Started with Litmus
-Check out the Litmus Docs to get started.
+Check out the Litmus Docs to get started.
## Contributing to Chaos Hub
Check out the Contributing Guidelines for the Chaos Hub
-## Things to Consider
-
-Some of the considerations that need to be made with Litmus (as a chaos framework), are broadly listed here. Many of these are already being worked on
-as mentioned in the [ROADMAP](./ROADMAP.md). For details or limitations around specific experiments, refer to the respective [experiments docs](https://docs.litmuschaos.io/docs/pod-delete/).
-
-- Litmus chaos operator and the chaos experiments run as kubernetes resources in the cluster. In case of airgapped environments, the chaos custom resources
- and images need to be hosted on premise.
-- When attempting to execute platform specific chaos experiments (like those on AWS, GCP cloud) the access details are passed via kubernetes secrets. Support
- for other modes of secret management with Litmus is yet to be tested/implemented.
-- Some chaos experiments make use of the docker api from within the experiment pods, and thereby require the docker socket to be mounted. User discretion is
- advised when allowing developers/devops admins/SREs access for running these experiments.
-- In (rare) cases where chaos experiments make use of privileged containers, the recommended security policies will be documented.
-
## Community
diff --git a/images/chaos-operator-flow.png b/images/chaos-operator-flow.png
new file mode 100644
index 00000000000..2d86ba9406f
Binary files /dev/null and b/images/chaos-operator-flow.png differ
diff --git a/images/litmus-control-and-execution-plane-overview.png b/images/litmus-control-and-execution-plane-overview.png
new file mode 100644
index 00000000000..40a62942e76
Binary files /dev/null and b/images/litmus-control-and-execution-plane-overview.png differ