diff --git a/bioindex/src/main/resources/compression.sh b/bioindex/src/main/resources/compression.sh
new file mode 100644
index 00000000..e143bb44
--- /dev/null
+++ b/bioindex/src/main/resources/compression.sh
@@ -0,0 +1,3 @@
+#!/bin/bash -xe
+
+sudo aws s3 cp s3://dig-data-registry/hail.jar /usr/lib/spark/jars/
diff --git a/bioindex/src/main/resources/geneExpression.py b/bioindex/src/main/resources/geneExpression.py
index 7f40f801..382df089 100644
--- a/bioindex/src/main/resources/geneExpression.py
+++ b/bioindex/src/main/resources/geneExpression.py
@@ -16,6 +16,7 @@ def main():
     # sort and write
     df.orderBy(['gene']) \
         .write \
+        .option("compression", "is.hail.io.compress.BGzipCodec") \
         .mode('overwrite') \
         .json(outdir)
 
diff --git a/bioindex/src/main/scala/GeneExpression.scala b/bioindex/src/main/scala/GeneExpressionStage.scala
similarity index 88%
rename from bioindex/src/main/scala/GeneExpression.scala
rename to bioindex/src/main/scala/GeneExpressionStage.scala
index 0f1e052e..e0042a5e 100644
--- a/bioindex/src/main/scala/GeneExpression.scala
+++ b/bioindex/src/main/scala/GeneExpressionStage.scala
@@ -1,8 +1,6 @@
 package org.broadinstitute.dig.aggregator.methods.bioindex
 
 import org.broadinstitute.dig.aggregator.core._
-import org.broadinstitute.dig.aws._
-import org.broadinstitute.dig.aws.config.emr._
 import org.broadinstitute.dig.aws.emr._
 
 /** The final result of all aggregator methods is building the BioIndex. All
@@ -21,7 +19,8 @@ class GeneExpressionStage(implicit context: Context) extends Stage {
 
   /** Use latest EMR release. */
   override val cluster: ClusterDef = super.cluster.copy(
-    releaseLabel = ReleaseLabel.emrLatest
+    releaseLabel = ReleaseLabel.emrLatest,
+    bootstrapScripts = Seq(new BootstrapScript(resourceUri("compression.sh")))
   )
 
   /** Output to Job steps. */