diff --git a/.scalafmt.conf b/.scalafmt.conf
index 458cf423..212bff49 100644
--- a/.scalafmt.conf
+++ b/.scalafmt.conf
@@ -10,4 +10,15 @@ runner.dialect = scala213source3
 spaces.beforeContextBoundColon = Always
 style = defaultWithAlign
 
-rewrite.scala3.convertToNewSyntax = true  # TODO remove when written in Scala3.
\ No newline at end of file
+rewrite.scala3.convertToNewSyntax = true  # TODO remove when written in Scala3.
+fileOverride {
+  "glob:**/scheduler-3/src/main/scala/**" {
+     runner.dialect = scala3
+  }
+  "glob:**/scheduler-3/target/scala-3.6.2/src_managed/main/**" {
+     runner.dialect = scala3
+  }
+  "glob:**/scheduler-3/src/test/scala/**" {
+     runner.dialect = scala3
+  }
+}
diff --git a/project/Dependencies.scala b/project/Dependencies.scala
index 70945677..4152bc51 100644
--- a/project/Dependencies.scala
+++ b/project/Dependencies.scala
@@ -34,6 +34,13 @@ object Dependencies {
     lazy val test                   = Seq(testKit, scalatest)
   }
 
+  object Circe {
+    private lazy val version = "0.14.10"
+
+    lazy val generic = "io.circe" %% "circe-generic" % version
+    lazy val parser = "io.circe" %% "circe-parser" % version
+  }
+
   object Fs2 {
     private lazy val version      = "3.11.0"
     private lazy val kafkaVersion = "3.6.0"
@@ -100,6 +107,10 @@ object Dependencies {
     lazy val testkit = "org.typelevel" %% "otel4s-oteljava-testkit" % version % Test
   }
 
+  object TopicLoader {
+    lazy val topicLoader = "uk.sky" %% "fs2-kafka-topic-loader" % "0.1.0"
+  }
+
   object Vulcan {
     private lazy val version = "1.11.1"
 
@@ -147,6 +158,7 @@ object Dependencies {
     scalaTest,
     scalaTestPlusMockito
   )
+
   val scheduler: Seq[ModuleID] = core ++ runtime ++ test
 
   val scheduler3: Seq[ModuleID] = Seq(
@@ -158,6 +170,8 @@ object Dependencies {
     Cats.effectTestkitScalatest,
     Cats.log4cats,
     Cats.log4catsSlf4j,
+    Circe.generic,
+    Circe.parser,
     Fs2.core,
     Fs2.kafka,
     Monocle.core,
@@ -171,6 +185,7 @@ object Dependencies {
     OpenTelemetry.sdkAutoconfigure,
     Otel4s.java,
     Otel4s.testkit,
-    mouse
+    mouse,
+    TopicLoader.topicLoader
   )
 }
diff --git a/scheduler-3/src/main/resources/application.conf b/scheduler-3/src/main/resources/application.conf
index a2eea4e6..40750153 100644
--- a/scheduler-3/src/main/resources/application.conf
+++ b/scheduler-3/src/main/resources/application.conf
@@ -1,5 +1,9 @@
 scheduler.reader {
   schedule-topics = [${?SCHEDULE_TOPICS}]
+  topics {
+    avro = [${?AVRO_SCHEDULE_TOPICS}]
+    json = [${?JSON_SCHEDULE_TOPICS}]
+  }
   kafka-brokers = "localhost:9092"
   kafka-brokers = ${?KAFKA_BROKERS}
 }
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/EventSubscriber.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/EventSubscriber.scala
index de1a9f7f..4210b4be 100644
--- a/scheduler-3/src/main/scala/uk/sky/scheduler/EventSubscriber.scala
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/EventSubscriber.scala
@@ -1,10 +1,141 @@
 package uk.sky.scheduler
 
-import fs2.Stream
+import cats.effect.Resource.ExitCase
+import cats.effect.kernel.Deferred
+import cats.effect.kernel.Resource.ExitCase.Succeeded
+import cats.effect.{Async, Ref, Resource}
+import cats.syntax.all.*
+import cats.{Monad, Parallel, Show}
+import fs2.{Stream, *}
+import fs2.kafka.*
+import mouse.all.*
+import org.typelevel.log4cats.LoggerFactory
+import org.typelevel.otel4s.metrics.Meter
+import org.typelevel.otel4s.{Attribute, Attributes}
+import uk.sky.fs2.kafka.topicloader.TopicLoader
+import uk.sky.scheduler.circe.jsonScheduleDecoder
+import uk.sky.scheduler.config.{Config, KafkaConfig}
 import uk.sky.scheduler.domain.ScheduleEvent
 import uk.sky.scheduler.error.ScheduleError
+import uk.sky.scheduler.kafka.avro.{avroBinaryDeserializer, avroScheduleCodec, AvroSchedule}
+import uk.sky.scheduler.kafka.json.{jsonDeserializer, JsonSchedule}
 import uk.sky.scheduler.message.Message
+import uk.sky.scheduler.converters.all.*
 
 trait EventSubscriber[F[_]] {
   def messages: Stream[F, Message[Either[ScheduleError, Option[ScheduleEvent]]]]
 }
+
+object EventSubscriber {
+  private type Output = Either[ScheduleError, Option[ScheduleEvent]]
+
+  def kafka[F[_] : Async : Parallel : LoggerFactory](
+      config: Config,
+      loaded: Deferred[F, Unit]
+  ): F[EventSubscriber[F]] = {
+
+    val avroConsumerSettings: ConsumerSettings[F, String, Either[ScheduleError, Option[AvroSchedule]]] = {
+      given Resource[F, Deserializer[F, Either[ScheduleError, Option[AvroSchedule]]]] =
+        avroBinaryDeserializer[F, AvroSchedule].map(_.option.map(_.sequence))
+
+      config.kafka.consumerSettings[F, String, Either[ScheduleError, Option[AvroSchedule]]]
+    }
+
+    val jsonConsumerSettings: ConsumerSettings[F, String, Either[ScheduleError, Option[JsonSchedule]]] = {
+      given Deserializer[F, Either[ScheduleError, Option[JsonSchedule]]] =
+        jsonDeserializer[F, JsonSchedule].option.map(_.sequence)
+
+      config.kafka.consumerSettings[F, String, Either[ScheduleError, Option[JsonSchedule]]]
+    }
+
+    for {
+      avroLoaderRef <- Ref.of[F, Boolean](false)
+      jsonLoadedRef <- Ref.of[F, Boolean](false)
+    } yield new EventSubscriber[F] {
+
+      /** If both topics have finished loading, complete the Deferred to allow Queueing schedules.
+        */
+      private def onLoadCompare(exitCase: ExitCase): F[Unit] = exitCase match {
+        case Succeeded                               =>
+          for {
+            avroLoaded <- avroLoaderRef.get
+            jsonLoaded <- jsonLoadedRef.get
+            _          <- Async[F].whenA(avroLoaded && jsonLoaded)(loaded.complete(()))
+          } yield ()
+        case ExitCase.Errored(_) | ExitCase.Canceled => Async[F].unit
+      }
+
+      private val avroStream: Stream[F, ConsumerRecord[String, Either[ScheduleError, Option[AvroSchedule]]]] =
+        config.scheduler.reader.topics.avro.toNel
+          .fold(Stream.exec(avroLoaderRef.set(true) *> onLoadCompare(ExitCase.Succeeded)))(
+            TopicLoader.loadAndRun(_, avroConsumerSettings) { exitCase =>
+              avroLoaderRef.set(true) *> onLoadCompare(exitCase)
+            }
+          )
+
+      private val jsonStream: Stream[F, ConsumerRecord[String, Either[ScheduleError, Option[JsonSchedule]]]] =
+        config.scheduler.reader.topics.json.toNel
+          .fold(Stream.exec(jsonLoadedRef.set(true) *> onLoadCompare(ExitCase.Succeeded)))(
+            TopicLoader.loadAndRun(_, jsonConsumerSettings) { exitCase =>
+              jsonLoadedRef.set(true) *> onLoadCompare(exitCase)
+            }
+          )
+
+      override def messages: Stream[F, Message[Output]] = avroStream.merge(jsonStream).map(p => p.toMessage)
+
+    }
+  }
+
+  def observed[F[_] : Monad : Parallel : LoggerFactory : Meter](delegate: EventSubscriber[F]): F[EventSubscriber[F]] = {
+    given Show[ScheduleError] = {
+      case _: ScheduleError.InvalidAvroError => "invalid-avro"
+      case _: ScheduleError.NotJsonError     => "not-json"
+      case _: ScheduleError.InvalidJsonError => "invalid-json"
+//      case _: ScheduleError.DecodeError => "decode"
+//      case _: ScheduleError.TransformationError => "transformation"
+    }
+
+    def updateAttributes(source: String) = Attributes(
+      Attribute("message.type", "update"),
+      Attribute("message.source", source)
+    )
+
+    def deleteAttributes(source: String, deleteType: String) = Attributes(
+      Attribute("message.type", "delete"),
+      Attribute("message.source", source),
+      Attribute("message.delete.type", deleteType)
+    )
+
+    def errorAttributes(source: String, error: ScheduleError) = Attributes(
+      Attribute("message.type", "error"),
+      Attribute("message.source", source),
+      Attribute("message.error.type", error.show)
+    )
+
+    for {
+      counter <- Meter[F].counter[Long]("event-subscriber").create
+      logger  <- LoggerFactory[F].create
+    } yield new EventSubscriber[F] {
+      override def messages: Stream[F, Message[Either[ScheduleError, Option[ScheduleEvent]]]] =
+        delegate.messages.evalTapChunk { case Message(key, source, value, metadata) =>
+          val logCtx = Map("key" -> key, "source" -> source)
+
+          value match {
+            case Right(Some(_)) =>
+              logger.info(logCtx)(show"Decoded UPDATE for [$key] from $source") &>
+                counter.inc(updateAttributes(source))
+            case Right(None)    =>
+              lazy val deleteType = metadata.isExpired.fold("expired", "canceled")
+              logger.info(logCtx)(show"Decoded DELETE type=[$deleteType] for [$key] from $source") &>
+                counter.inc(deleteAttributes(source, deleteType))
+            case Left(error)    =>
+              logger.error(logCtx, error)(show"Error decoding [$key] from $source") &>
+                counter.inc(errorAttributes(source, error))
+          }
+        }
+    }
+  }
+
+  def live[F[_] : Async : Parallel : LoggerFactory : Meter](config: Config, loaded: Deferred[F, Unit]) =
+    kafka[F](config, loaded).flatMap(observed)
+}
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/circe/serdes.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/circe/serdes.scala
new file mode 100644
index 00000000..ef6196c8
--- /dev/null
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/circe/serdes.scala
@@ -0,0 +1,19 @@
+package uk.sky.scheduler.circe
+
+import uk.sky.scheduler.kafka.json.JsonSchedule
+import io.circe.*
+import io.circe.generic.semiauto
+import cats.syntax.all.*
+
+given jsonScheduleDecoder: Decoder[JsonSchedule] =
+  Decoder.forProduct5[JsonSchedule, Long, String, String, Option[String], Option[Map[String, String]]](
+    "time",
+    "topic",
+    "key",
+    "value",
+    "headers"
+  ) { (time, topic, key, value, headers) =>
+    JsonSchedule(time, topic, key, value, headers.combineAll)
+  }
+
+given scheduleEncoder: Encoder[JsonSchedule] = semiauto.deriveEncoder[JsonSchedule]
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/config/Config.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/config/Config.scala
index 20fd6b30..4d2bca7b 100644
--- a/scheduler-3/src/main/scala/uk/sky/scheduler/config/Config.scala
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/config/Config.scala
@@ -59,6 +59,32 @@ final case class SchedulerConfig(
 ) derives ConfigReader
 
 final case class ReaderConfig(
+    @deprecated("Remove in favour of 'topics'")
     scheduleTopics: List[String],
+    topics: TopicConfig,
     kafkaBrokers: String
 ) derives ConfigReader
+
+//object ReaderConfig {
+//  given readerConfigReader: ConfigReader[ReaderConfig] =
+//    ConfigReader
+//      .forProduct3[ReaderConfig, List[String], TopicConfig, String]("scheduleTopics", "topics", "kafkaBrokers")(
+//        ReaderConfig.apply
+//      )
+//      .ensure(
+//        readerConfig => readerConfig.scheduleTopics.nonEmpty,
+//        message = _ => "Schedule topics are empty"
+//      )
+//}
+
+final case class TopicConfig(avro: List[String], json: List[String])
+
+object TopicConfig {
+  given topicConfigReader: ConfigReader[TopicConfig] =
+    ConfigReader
+      .forProduct2[TopicConfig, List[String], List[String]]("avro", "json")(TopicConfig.apply)
+      .ensure(
+        config => config.avro.nonEmpty || config.json.nonEmpty,
+        _ => "Both Avro and JSON topics were empty"
+      )
+}
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/converters/all.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/converters/all.scala
new file mode 100644
index 00000000..5f4001fe
--- /dev/null
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/converters/all.scala
@@ -0,0 +1,3 @@
+package uk.sky.scheduler.converters
+
+object all extends ConsumerRecordConverter
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/converters/consumerRecord.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/converters/consumerRecord.scala
new file mode 100644
index 00000000..6f54d212
--- /dev/null
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/converters/consumerRecord.scala
@@ -0,0 +1,16 @@
+package uk.sky.scheduler.converters
+
+import fs2.kafka.ConsumerRecord
+import uk.sky.scheduler.domain.ScheduleEvent
+import uk.sky.scheduler.error.ScheduleError
+import uk.sky.scheduler.kafka.avro.AvroSchedule
+import uk.sky.scheduler.kafka.json.JsonSchedule
+import uk.sky.scheduler.message.Message
+
+private trait ConsumerRecordConverter {
+  extension (cr: ConsumerRecord[String, Either[ScheduleError, Option[JsonSchedule | AvroSchedule]]]) {
+    def toMessage: Message[Either[ScheduleError, Option[ScheduleEvent]]] = ???
+  }
+}
+
+object consumerRecord extends ConsumerRecordConverter
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/error/ScheduleError.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/error/ScheduleError.scala
index dc499c17..01c60f43 100644
--- a/scheduler-3/src/main/scala/uk/sky/scheduler/error/ScheduleError.scala
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/error/ScheduleError.scala
@@ -7,6 +7,12 @@ import org.apache.avro.Schema
 enum ScheduleError(val message: String, val cause: Throwable) extends Throwable(message, cause) {
   case InvalidAvroError(schema: Schema, error: Throwable)
       extends ScheduleError(show"Avro message did not conform to Schema: ${schema.getFullName}: $schema", error)
+
+  case NotJsonError(payload: String, error: Throwable) extends ScheduleError(s"'$payload' was not valid JSON", error)
+
+  case InvalidJsonError(payload: String, error: Throwable)
+      extends ScheduleError(s"JSON '$payload' did not conform to Schema", error)
+
 }
 
 object ScheduleError {
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/avro/AvroSchedule.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/avro/AvroSchedule.scala
new file mode 100644
index 00000000..2153ae84
--- /dev/null
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/avro/AvroSchedule.scala
@@ -0,0 +1,11 @@
+package uk.sky.scheduler.kafka.avro
+
+final case class AvroSchedule(
+    time: Long,
+    topic: String,
+    key: Array[Byte],
+    value: Option[Array[Byte]],
+    optionalHeaders: Option[Map[String, Array[Byte]]]
+) {
+  val headers: Map[String, Array[Byte]] = optionalHeaders.getOrElse(Map.empty[String, Array[Byte]])
+}
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/avro/serdes.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/avro/serdes.scala
new file mode 100644
index 00000000..dd684974
--- /dev/null
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/avro/serdes.scala
@@ -0,0 +1,45 @@
+package uk.sky.scheduler.kafka.avro
+
+import cats.effect.Sync
+import cats.effect.kernel.Resource
+import cats.syntax.all.*
+import fs2.kafka.{Deserializer, Serializer, ValueDeserializer, ValueSerializer}
+import org.apache.avro.Schema
+import uk.sky.scheduler.error.ScheduleError
+import vulcan.Codec
+
+given avroScheduleCodec: Codec[AvroSchedule] = Codec.record[AvroSchedule](
+  name = "ScheduleWithHeaders",
+  namespace = "com.sky.kms.domain.Schedule"
+)(field =>
+  (
+    field("time", _.time, doc = "The time to execute the Schedule, in epoch milliseconds.".some),
+    field("topic", _.topic, doc = "The topic to send the Schedule to.".some),
+    field("key", _.key, doc = "The key identifying the payload.".some),
+    field("value", _.value, doc = "The payload to be sent. null indicates a tombstone.".some),
+    field("headers", _.optionalHeaders, doc = "Optional extra metadata to send with the payload.".some)
+  ).mapN(AvroSchedule.apply)
+)
+
+def avroBinaryDeserializer[F[_] : Sync, V : Codec]: Resource[F, ValueDeserializer[F, Either[ScheduleError, V]]] =
+  Codec[V].schema match {
+    case Left(error)   => Resource.raiseError(error.throwable)
+    case Right(schema) => Resource.pure(avroBinaryDeserializer(schema))
+  }
+
+def avroBinaryDeserializer[F[_] : Sync, V : Codec](schema: Schema): ValueDeserializer[F, Either[ScheduleError, V]] =
+  Deserializer.lift[F, Either[ScheduleError, V]](bytes =>
+    Sync[F].delay(
+      Codec
+        .fromBinary[V](bytes, schema)
+        .leftMap(e => ScheduleError.InvalidAvroError(schema, e.throwable))
+    )
+  )
+
+def avroBinarySerializer[F[_] : Sync, V : Codec]: ValueSerializer[F, V] =
+  Serializer.lift[F, V](
+    Codec
+      .toBinary[V](_)
+      .leftMap(_.throwable)
+      .liftTo[F]
+  )
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/json/JsonSchedule.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/json/JsonSchedule.scala
new file mode 100644
index 00000000..ea73b012
--- /dev/null
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/json/JsonSchedule.scala
@@ -0,0 +1,9 @@
+package uk.sky.scheduler.kafka.json
+
+final case class JsonSchedule(
+    time: Long,
+    topic: String,
+    key: String,
+    value: Option[String],
+    headers: Map[String, String]
+)
diff --git a/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/json/serdes.scala b/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/json/serdes.scala
new file mode 100644
index 00000000..cd7a220d
--- /dev/null
+++ b/scheduler-3/src/main/scala/uk/sky/scheduler/kafka/json/serdes.scala
@@ -0,0 +1,19 @@
+package uk.sky.scheduler.kafka.json
+
+import cats.effect.Sync
+import cats.syntax.all.*
+import io.circe.syntax.*
+import fs2.kafka.{Deserializer, Serializer, ValueDeserializer, ValueSerializer}
+import io.circe.{parser, Decoder, Encoder}
+import uk.sky.scheduler.error.ScheduleError
+
+def jsonDeserializer[F[_] : Sync, T : Decoder]: ValueDeserializer[F, Either[ScheduleError, T]] =
+  for {
+    payload <- Deserializer.string[F]
+  } yield for {
+    json    <- parser.parse(payload).leftMap(ScheduleError.NotJsonError(payload, _))
+    decoded <- json.as[T].leftMap(ScheduleError.InvalidJsonError(json.noSpaces, _))
+  } yield decoded
+
+def jsonSerializer[F[_] : Sync, V : Encoder]: ValueSerializer[F, V] =
+  Serializer.string[F].contramap[V](_.asJson.noSpaces)