diff --git a/ci-operator/config/openshift/installer/master.yaml b/ci-operator/config/openshift/installer/master.yaml
index f23b3d18e837b..a597e7caabb52 100644
--- a/ci-operator/config/openshift/installer/master.yaml
+++ b/ci-operator/config/openshift/installer/master.yaml
@@ -4,6 +4,11 @@ base_images:
     name: origin-v4.0
     namespace: openshift
     tag: base
+  base-smoke:
+    cluster: https://api.ci.openshift.org
+    namespace: openshift
+    name: origin-release
+    tag: bazel
 binary_build_commands: go build ./installer/cmd/tectonic
 canonical_go_repository: github.com/openshift/installer
 images:
@@ -18,6 +23,20 @@ images:
       as:
       - build
   to: installer
+- dockerfile_path: images/tectonic-installer/Dockerfile.ci
+  from: base-smoke
+  inputs:
+    bin:
+      paths:
+      - destination_dir: .
+        source_path: /go/src/github.com/openshift/installer/tectonic
+    root:
+      as:
+      - build
+  to: installer-bazel
+- dockerfile_path: images/installer-origin-release/Dockerfile.ci
+  from: installer-bazel
+  to: installer-smoke
 resources:
   '*':
     limits:
diff --git a/ci-operator/jobs/openshift/installer/openshift-installer-presubmits.yaml b/ci-operator/jobs/openshift/installer/openshift-installer-presubmits.yaml
index 127f3b459d440..4ce881026726b 100644
--- a/ci-operator/jobs/openshift/installer/openshift-installer-presubmits.yaml
+++ b/ci-operator/jobs/openshift/installer/openshift-installer-presubmits.yaml
@@ -90,6 +90,62 @@ presubmits:
           - configMap:
               name: cluster-profile-aws
     trigger: ((?m)^/test( all| e2e-aws),?(\s+|$))
+  - agent: kubernetes
+    always_run: false
+    branches:
+    - master
+    context: ci/prow/e2e-aws-smoke
+    decorate: true
+    name: pull-ci-origin-installer-e2e-aws-smoke
+    rerun_command: /test e2e-aws-smoke
+    # The abomination below is equivalent to `^((?!Documentation).)*$`. Since
+    # Go doesn't support negative lookaheads, we are stuck with the following.
+    run_if_changed: ^([^D]|D(D|oD|ocD|ocuD|ocum(D|e(D|n(D|t(D|aD|atD|atiD|atioD)))))*([^Do]|o[^Dc]|oc[^Du]|ocu[^Dm]|ocum([^De]|e([^Dn]|n([^Dt]|t([^Da]|a[^Dt]|at[^Di]|ati[^Do]|atio[^Dn]))))))*(D(D|oD|ocD|ocuD|ocum(D|e(D|n(D|t(D|aD|atD|atiD|atioD)))))*(o|oc|ocu|ocum(e(n(t(a|at|ati|atio)?)?)?)?)?)?$
+    skip_cloning: true
+    spec:
+      containers:
+      - command:
+        - ci-operator
+        - --artifact-dir=$(ARTIFACTS)
+        - --secret-dir=/usr/local/e2e-aws-smoke-cluster-profile
+        - --template=/usr/local/e2e-aws-smoke
+        - --target=e2e-aws-smoke
+        env:
+        - name: JOB_NAME_SAFE
+          value: e2e-aws-smoke
+        - name: CLUSTER_TYPE
+          value: aws
+        - name: CONFIG_SPEC
+          valueFrom:
+            configMapKeyRef:
+              key: master.json
+              name: ci-operator-openshift-installer
+        image: ci-operator:latest
+        name: test
+        resources:
+          limits:
+            cpu: 500m
+          requests:
+            cpu: 10m
+        volumeMounts:
+        - mountPath: /usr/local/e2e-aws-smoke
+          name: job-definition
+          subPath: cluster-launch-installer-e2e-smoke.yaml
+        - mountPath: /usr/local/e2e-aws-smoke-cluster-profile
+          name: cluster-profile
+      serviceAccountName: ci-operator
+      volumes:
+      - configMap:
+          name: prow-job-cluster-launch-installer-e2e
+        name: job-definition
+      - name: cluster-profile
+        projected:
+          sources:
+          - secret:
+              name: cluster-secrets-aws
+          - configMap:
+              name: cluster-profile-aws
+    trigger: ((?m)^/test( all| e2e-aws-smoke),?(\s+|$))
   - agent: kubernetes
     always_run: false
     branches:
diff --git a/ci-operator/templates/cluster-launch-installer-e2e-smoke.yaml b/ci-operator/templates/cluster-launch-installer-e2e-smoke.yaml
new file mode 100644
index 0000000000000..527bd7bffa874
--- /dev/null
+++ b/ci-operator/templates/cluster-launch-installer-e2e-smoke.yaml
@@ -0,0 +1,342 @@
+kind: Template
+apiVersion: template.openshift.io/v1
+
+parameters:
+- name: JOB_NAME_SAFE
+  required: true
+- name: JOB_NAME_HASH
+  required: true
+- name: NAMESPACE
+  required: true
+- name: IMAGE_CLI
+  required: true
+- name: IMAGE_FORMAT
+  required: true
+- name: IMAGE_INSTALLER
+  required: true
+- name: IMAGE_TESTS
+  required: true
+- name: IMAGE_INSTALLER_SMOKE
+  required: true
+- name: CLUSTER_TYPE
+  required: true
+
+objects:
+
+# We want the cluster to be able to access these images
+- kind: RoleBinding
+  apiVersion: authorization.openshift.io/v1
+  metadata:
+    name: ${JOB_NAME_SAFE}-image-puller
+    namespace: ${NAMESPACE}
+  roleRef:
+    name: system:image-puller
+  subjects:
+  - kind: SystemGroup
+    name: system:unauthenticated
+
+# The e2e pod spins up a cluster, runs e2e tests, and then cleans up the cluster.
+- kind: Pod
+  apiVersion: v1
+  metadata:
+    name: ${JOB_NAME_SAFE}
+    namespace: ${NAMESPACE}
+    annotations:
+      # we want to gather the teardown logs no matter what
+      ci-operator.openshift.io/wait-for-container-artifacts: teardown
+  spec:
+    restartPolicy: Never
+    activeDeadlineSeconds: 10800
+    terminationGracePeriodSeconds: 900
+    volumes:
+    - name: artifacts
+      emptyDir: {}
+    - name: shared-tmp
+      emptyDir: {}
+    - name: cluster-profile
+      secret:
+        secretName: ${JOB_NAME_SAFE}-cluster-profile
+
+    initContainers:
+    - name: cli
+      image: ${IMAGE_CLI}
+      volumeMounts:
+      - name: shared-tmp
+        mountPath: /tmp/shared
+      command:
+      - cp
+      - /usr/bin/oc
+      - /tmp/shared/oc
+
+    - name: smoke-test
+      image: ${IMAGE_INSTALLER_SMOKE}
+      volumeMounts:
+      - name: shared-tmp
+        mountPath: /tmp/shared
+      command:
+      - cp
+      - /usr/bin/smoke
+      - /tmp/shared/smoke
+
+    containers:
+
+    # Once admin.kubeconfig exists, executes shared tests
+    - name: test
+      image: ${IMAGE_INSTALLER_SMOKE}
+      resources:
+        requests:
+          cpu: 1
+          memory: 300Mi
+        limits:
+          cpu: 3
+          memory: 2Gi
+      volumeMounts:
+      - name: shared-tmp
+        mountPath: /tmp/shared
+      - name: cluster-profile
+        mountPath: /tmp/cluster
+      - name: artifacts
+        mountPath: /tmp/artifacts
+      env:
+      - name: HOME
+        value: /tmp/home
+      - name: USER
+        value: bazel
+      - name: KUBECONFIG
+        value: /tmp/admin.kubeconfig
+      command:
+      - /bin/bash
+      - -c
+      - |
+        #!/bin/bash
+        set -euo pipefail
+
+        trap 'touch /tmp/shared/exit' EXIT
+        trap 'kill $(jobs -p); exit 0' TERM
+
+        mkdir -p "${HOME}"
+
+        # wait until the setup job creates admin.kubeconfig
+        while true; do
+          if [[ -f /tmp/shared/exit ]]; then
+            echo "Another process exited" 2>&1
+            exit 1
+          fi
+          if [[ ! -f /tmp/shared/admin.kubeconfig ]]; then
+            sleep 15 & wait
+            continue
+          fi
+          break
+        done
+        echo "Found shared kubeconfig"
+
+        # don't let clients impact the global kubeconfig
+        cp /tmp/shared/admin.kubeconfig /tmp/admin.kubeconfig
+
+        PATH=/usr/libexec/origin:$PATH
+
+        # set up cloud provider specific env vars
+        if [[ "${CLUSTER_TYPE}" == "gcp" ]]; then
+          export GOOGLE_APPLICATION_CREDENTIALS="/tmp/cluster/gce.json"
+          export KUBE_SSH_USER=cloud-user
+          mkdir -p ~/.ssh
+          cp /tmp/cluster/ssh-privatekey ~/.ssh/google_compute_engine || true
+          export PROVIDER_ARGS='-provider=gce -gce-zone=us-east1-c -gce-project=openshift-gce-devel-ci'
+        elif [[ "${CLUSTER_TYPE}" == "aws" ]]; then
+          region="$( python -c 'import sys, json; print json.load(sys.stdin)["tectonic_aws_region"]' </tmp/shared/cluster/terraform.tfvars )"
+          export PROVIDER_ARGS="-provider=aws -gce-zone=${region}"
+        fi
+
+        mkdir -p /tmp/output
+        cd /tmp/output
+
+        exit ${rc:-0}
+
+    # Runs an install
+    - name: setup
+      image: ${IMAGE_INSTALLER_SMOKE}
+      volumeMounts:
+      - name: shared-tmp
+        mountPath: /tmp
+      - name: cluster-profile
+        mountPath: /etc/openshift-installer
+      - name: artifacts
+        mountPath: /tmp/artifacts
+      env:
+      - name: INSTANCE_PREFIX
+        value: ${NAMESPACE}-${JOB_NAME_HASH}
+      - name: TYPE
+        value: ${CLUSTER_TYPE}
+      - name: AWS_SHARED_CREDENTIALS_FILE
+        value: /etc/openshift-installer/.awscred
+      - name: KUBECONFIG
+        value: /tmp/cluster/generated/auth/kubeconfig
+      command:
+      - /bin/bash
+      - -c
+      - |
+        #!/bin/bash
+        set -euo pipefail
+
+        trap 'rc=$?; if [[ $rc -ne 0 ]]; then touch /tmp/exit; fi; exit $rc' EXIT
+        trap 'kill $(jobs -p); exit 0' TERM
+
+        mkdir /tmp/cluster
+        cp /etc/openshift-installer/* /tmp/cluster/
+
+        export NAME=${INSTANCE_PREFIX}
+        (
+          export EMAIL=test@ci.openshift.io
+          export PASSWORD=$( date +%s | sha256sum | base64 | head -c 32 ; echo )
+          export SSH_KEY="$( cat /etc/openshift-installer/ssh-publickey )"
+          export EC2_AMI_OVERRIDE=ami-07307c397daf4d02e
+          export EXPIRATION_DATE=$( date -d '4 hours' --iso=minutes --utc )
+          envsubst </etc/openshift-installer/openshift.yaml >/tmp/cluster/inputs.yaml
+        )
+        mkdir /tmp/artifacts/installer
+        cp /tmp/cluster/inputs.yaml /tmp/artifacts/installer/
+
+        echo "Invoking installer ..."
+
+        cd /tmp/cluster
+        tectonic init --config=inputs.yaml
+        mv -f ${NAME}/* /tmp/cluster/
+        cp config.yaml internal.yaml terraform.tfvars /tmp/artifacts/installer/
+
+        tectonic install --dir=. --log-level=debug
+
+        # wait until oc shows up
+        while true; do
+          if [[ -f /tmp/exit ]]; then
+            echo "Interrupted"
+            cp $KUBECONFIG /tmp/admin.kubeconfig
+            exit 1
+          fi
+          if [[ ! -f /tmp/oc ]]; then
+            echo "Waiting for oc binary to show up ..."
+            sleep 15 & wait
+            continue
+          fi
+          if ! /tmp/oc get nodes 2>/dev/null; then
+            echo "Waiting for API at $(/tmp/oc whoami --show-server) to respond ..."
+            sleep 15 & wait
+            continue
+          fi
+          # check multiple namespaces while we are transitioning to the new locations
+          if /tmp/oc get deploy/router -n tectonic-ingress 2>/dev/null; then
+            router_namespace=tectonic-ingress
+          elif /tmp/oc get deploy/router -n openshift-ingress 2>/dev/null; then
+            router_namespace=openshift-ingress
+          elif /tmp/oc get deploy/router -n default 2>/dev/null; then
+            router_namespace=default
+          else
+            echo "Waiting for router to be created ..."
+            sleep 15 & wait
+            continue
+          fi
+          break
+        done
+        if ! /tmp/oc wait deploy/router -n "${router_namespace}" --for condition=available --timeout=20m; then
+          echo "Installation failed"
+          cp $KUBECONFIG /tmp/admin.kubeconfig
+          exit 1
+        fi
+        cp $KUBECONFIG /tmp/admin.kubeconfig
+        echo "Installation successful"
+        echo "Starting installer smoke tests..."
+        export SMOKE_KUBECONFIG=${KUBECONFIG}
+        export SMOKE_MANIFEST_PATHS=/tmp/cluster/generated/manifests
+        # 3 masters/3 workers/1 bootstrap
+        export SMOKE_NODE_COUNT=7
+        if ! /tmp/smoke -cluster -test.v; then
+          echo "Smoke tests failed"
+          exit 1
+        fi
+        echo "Smoke tests passed"
+
+    # Performs cleanup of all created resources
+    - name: teardown
+      image: ${IMAGE_INSTALLER_SMOKE}
+      volumeMounts:
+      - name: shared-tmp
+        mountPath: /tmp/shared
+      - name: cluster-profile
+        mountPath: /etc/openshift-installer
+      - name: artifacts
+        mountPath: /tmp/artifacts
+      env:
+      - name: INSTANCE_PREFIX
+        value: ${NAMESPACE}-${JOB_NAME_HASH}
+      - name: TYPE
+        value: ${CLUSTER_TYPE}
+      - name: KUBECONFIG
+        value: /tmp/shared/admin.kubeconfig
+      command:
+      - /bin/bash
+      - -c
+      - |
+        #!/bin/bash
+        function teardown() {
+          set +e
+          touch /tmp/shared/exit
+          export PATH=$PATH:/tmp/shared
+
+          echo "Gathering artifacts ..."
+          mkdir -p /tmp/artifacts/pods /tmp/artifacts/nodes /tmp/artifacts/metrics
+
+          oc --request-timeout=5s get nodes -o jsonpath --template '{range .items[*]}{.metadata.name}{"\n"}{end}' > /tmp/nodes
+          oc --request-timeout=5s get pods --all-namespaces --template '{{ range .items }}{{ $name := .metadata.name }}{{ $ns := .metadata.namespace }}{{ range .spec.containers }}-n {{ $ns }} {{ $name }} -c {{ .name }}{{ "\n" }}{{ end }}{{ range .spec.initContainers }}-n {{ $ns }} {{ $name }} -c {{ .name }}{{ "\n" }}{{ end }}{{ end }}' > /tmp/containers
+          oc --request-timeout=5s get nodes -o json > /tmp/artifacts/nodes.json
+          oc --request-timeout=5s get events --all-namespaces -o json > /tmp/artifacts/events.json
+          oc --request-timeout=5s get pods -l openshift.io/component=api --all-namespaces --template '{{ range .items }}-n {{ .metadata.namespace }} {{ .metadata.name }}{{ "\n" }}{{ end }}' > /tmp/pods-api
+
+          # gather nodes first in parallel since they may contain the most relevant debugging info
+          while IFS= read -r i; do
+            mkdir -p /tmp/artifacts/nodes/$i
+            (
+            oc get --request-timeout=20s --raw /api/v1/nodes/$i/proxy/logs/messages | gzip -c > /tmp/artifacts/nodes/$i/messages.gz
+            oc get --request-timeout=20s --raw /api/v1/nodes/$i/proxy/logs/journal | sed -e 's|.*href="\(.*\)".*|\1|;t;d' > /tmp/journals
+            while IFS= read -r j; do
+              oc get --request-timeout=20s --raw /api/v1/nodes/$i/proxy/logs/journal/${j}system.journal | gzip -c > /tmp/artifacts/nodes/$i/journal.gz
+            done < /tmp/journals
+            oc get --request-timeout=20s --raw /api/v1/nodes/$i/proxy/metrics | gzip -c > /tmp/artifacts/metrics/node-$i.gz
+            oc get --request-timeout=20s --raw /api/v1/nodes/$i/proxy/debug/pprof/heap > /tmp/artifacts/nodes/$i/heap
+            oc get --request-timeout=20s --raw /api/v1/nodes/$i/proxy/logs/secure | gzip -c > /tmp/artifacts/nodes/$i/secure.gz
+            oc get --request-timeout=20s --raw /api/v1/nodes/$i/proxy/logs/audit | gzip -c > /tmp/artifacts/nodes/$i/audit.gz
+            ) &
+          done < /tmp/nodes
+
+          while IFS= read -r i; do
+            file="$( echo "$i" | cut -d ' ' -f 3 | tr -s ' ' '_' )"
+            oc exec $i -- /bin/bash -c 'oc get --raw /debug/pprof/heap --server "https://$( hostname ):8443" --config /etc/origin/master/admin.kubeconfig' > /tmp/artifacts/metrics/${file}-heap
+            oc exec $i -- /bin/bash -c 'oc get --raw /metrics --server "https://$( hostname ):8443" --config /etc/origin/master/admin.kubeconfig' | gzip -c > /tmp/artifacts/metrics/${file}-api.gz
+            oc exec $i -- /bin/bash -c 'oc get --raw /debug/pprof/heap --server "https://$( hostname ):8444" --config /etc/origin/master/admin.kubeconfig' > /tmp/artifacts/metrics/${file}-controllers-heap
+            oc exec $i -- /bin/bash -c 'oc get --raw /metrics --server "https://$( hostname ):8444" --config /etc/origin/master/admin.kubeconfig' | gzip -c > /tmp/artifacts/metrics/${file}-controllers.gz
+          done < /tmp/pods-api
+
+          while IFS= read -r i; do
+            file="$( echo "$i" | cut -d ' ' -f 2,3,5 | tr -s ' ' '_' )"
+            oc logs --request-timeout=20s $i | gzip -c > /tmp/artifacts/pods/${file}.log.gz
+            oc logs --request-timeout=20s -p $i | gzip -c > /tmp/artifacts/pods/${file}_previous.log.gz
+          done < /tmp/containers
+
+          echo "Waiting for node logs to finish ..."
+          wait
+
+          echo "Deprovisioning cluster ..."
+          export AWS_SHARED_CREDENTIALS_FILE=/etc/openshift-installer/.awscred
+          cd /tmp/shared/cluster
+          for i in `seq 1 2`; do
+            tectonic destroy --dir=. --log-level=debug
+          done
+        }
+
+        trap 'teardown' EXIT
+        trap 'kill $(jobs -p); exit 0' TERM
+
+        for i in `seq 1 120`; do
+          if [[ -f /tmp/shared/exit ]]; then
+            exit 0
+          fi
+          sleep 60 & wait
+        done
diff --git a/projects/origin-release/bazel/Dockerfile b/projects/origin-release/bazel/Dockerfile
new file mode 100644
index 0000000000000..03ea11bcb2ef3
--- /dev/null
+++ b/projects/origin-release/bazel/Dockerfile
@@ -0,0 +1,11 @@
+# This Dockerfile is used to create origin-release:bazel necessary to build
+# openshift/installer smoke tests.
+
+FROM docker.io/openshift/origin-release:golang-1.10
+
+RUN wget -P /etc/yum.repos.d https://copr.fedorainfracloud.org/coprs/vbatts/bazel/repo/epel-7/vbatts-bazel-epel-7.repo
+
+RUN yum install -y bazel gcc-c++ libtool && \
+    yum clean all
+
+WORKDIR /go/src/github.com/openshift/origin
diff --git a/projects/origin-release/pipeline.yaml b/projects/origin-release/pipeline.yaml
index b94e875851f0f..2585e67d397b9 100644
--- a/projects/origin-release/pipeline.yaml
+++ b/projects/origin-release/pipeline.yaml
@@ -161,6 +161,36 @@ objects:
     triggers:
     - imageChange: {}
       type: ImageChange
+- apiVersion: v1
+  kind: BuildConfig
+  metadata:
+    labels:
+      app: origin-release
+    name: "origin-release-bazel"
+  spec:
+    successfulBuildsHistoryLimit: 1
+    failedBuildsHistoryLimit: 2
+    output:
+      to:
+        kind: DockerImage
+        name: "${PUSH_REPOSITORY}:bazel"
+      pushSecret:
+        name: "dockerhub"
+    source:
+      git:
+        ref: "${SOURCE_REF}"
+        uri: "${SOURCE_URL}"
+      contextDir: "projects/origin-release/bazel/"
+      type: Git
+    strategy:
+      dockerStrategy:
+        from:
+          kind: ImageStreamTag
+          name: origin-release:base
+        imageOptimizationPolicy: SkipLayers
+    triggers:
+    - imageChange: {}
+      type: ImageChange
 - apiVersion: v1
   kind: BuildConfig
   metadata: