diff --git a/services/trading_service/src/services/enhanced_ml.rs b/services/trading_service/src/services/enhanced_ml.rs
index 9b77d8e3b..3cfe22dfa 100644
--- a/services/trading_service/src/services/enhanced_ml.rs
+++ b/services/trading_service/src/services/enhanced_ml.rs
@@ -23,8 +23,9 @@ use serde::{Deserialize, Serialize};
 use std::collections::HashMap;
 use std::sync::Arc;
 use std::time::{Instant, SystemTime, UNIX_EPOCH};
-use tokio::sync::{broadcast, RwLock};
+use tokio::sync::{broadcast, OnceCell, RwLock};
 use tokio_stream::{wrappers::BroadcastStream, wrappers::ReceiverStream, Stream, StreamExt};
+use tonic::transport::Channel;
 use tonic::{Request, Response, Status};
 use tracing::{debug, info, warn, instrument};
 
@@ -189,6 +190,8 @@ pub struct EnhancedMLServiceImpl {
     feature_preprocessor: Arc<FeaturePreprocessor>,
     // System metrics tracking
     system: Arc<RwLock<System>>,
+    // gRPC client for forwarding retrain requests to ml_training_service
+    ml_training_client: Arc<OnceCell<crate::proto::ml_training::ml_training_service_client::MlTrainingServiceClient<Channel>>>,
 }
 
 impl EnhancedMLServiceImpl {
@@ -210,9 +213,29 @@ impl EnhancedMLServiceImpl {
             ml_fallback_manager,
             feature_preprocessor: Arc::new(FeaturePreprocessor::new()),
             system: Arc::new(RwLock::new(System::new_all())),
+            ml_training_client: Arc::new(OnceCell::new()),
         }
     }
 
+    /// Lazily connect to ml_training_service. Endpoint from ML_TRAINING_SERVICE_URL env var.
+    async fn get_training_client(
+        &self,
+    ) -> Result<crate::proto::ml_training::ml_training_service_client::MlTrainingServiceClient<Channel>, Status> {
+        self.ml_training_client
+            .get_or_try_init(|| async {
+                let url = std::env::var("ML_TRAINING_SERVICE_URL")
+                    .unwrap_or_else(|_| "http://ml-training-service:50052".to_string());
+                let channel = Channel::from_shared(url)
+                    .map_err(|e| Status::internal(format!("invalid training service URL: {e}")))?
+                    .connect()
+                    .await
+                    .map_err(|e| Status::unavailable(format!("ml_training_service unreachable: {e}")))?;
+                Ok(crate::proto::ml_training::ml_training_service_client::MlTrainingServiceClient::new(channel))
+            })
+            .await
+            .cloned()
+    }
+
     /// Load model from checkpoint file (production implementation)
     pub async fn load_model_from_file(
         &self,
@@ -1020,16 +1043,51 @@ impl MlService for EnhancedMLServiceImpl {
         request: Request<RetrainModelRequest>,
     ) -> Result<Response<RetrainModelResponse>, Status> {
         let req = request.into_inner();
+        info!(model_name = %req.model_name, "retrain_model: forwarding to ml_training_service");
 
-        warn!(
-            model_name = %req.model_name,
-            "retrain_model called but training service integration not yet implemented"
-        );
+        // Look up model to validate it exists and get its ModelType
+        let models = self.models.read().await;
+        let model_info = models.get(&req.model_name).ok_or_else(|| {
+            Status::not_found(format!("model '{}' not registered", req.model_name))
+        })?;
+        let model_type_str = model_info.model_type.as_str().to_string();
+        drop(models);
 
-        Err(Status::unavailable(format!(
-            "Model retraining for '{}' not yet wired to ml_training_service",
-            req.model_name
-        )))
+        // Connect to ml_training_service
+        let mut client = self.get_training_client().await?;
+
+        // Build fine-tune request
+        use crate::proto::ml_training::{StartTrainingRequest as MlStartReq, TrainingMode};
+        let training_req = MlStartReq {
+            model_type: model_type_str,
+            data_source: None,
+            hyperparameters: None,
+            use_gpu: true,
+            description: format!("Fine-tune {} via retrain_model RPC", req.model_name),
+            tags: req.parameters.clone(),
+            mode: TrainingMode::FineTune.into(),
+            resume_checkpoint_path: String::new(),
+            max_epochs: 10,
+        };
+
+        let resp = client
+            .start_training(tonic::Request::new(training_req))
+            .await
+            .map_err(|e| {
+                warn!(error = %e, "ml_training_service StartTraining failed");
+                Status::unavailable(format!("training service error: {e}"))
+            })?
+            .into_inner();
+
+        Ok(Response::new(RetrainModelResponse {
+            success: true,
+            message: resp.message,
+            job_id: Some(resp.job_id),
+            started_at: std::time::SystemTime::now()
+                .duration_since(std::time::UNIX_EPOCH)
+                .map(|d| d.as_secs() as i64)
+                .unwrap_or(0),
+        }))
     }
 
     #[instrument(skip_all)]
@@ -2769,24 +2827,24 @@ mod enhanced_ml_tests {
     }
 
     // -----------------------------------------------------------------------
-    // 12. retrain_model returns Unimplemented (H7 fix)
+    // 12. retrain_model returns NotFound for unknown models
     // -----------------------------------------------------------------------
     #[test]
-    fn test_retrain_model_status_code_is_unimplemented() {
-        // Verify the Status::unimplemented message format matches implementation
-        let model_name = "test-ppo-v2";
-        let status = Status::unimplemented(format!(
-            "Model retraining for '{}' not yet connected to training service",
+    fn test_retrain_model_not_found_for_unknown_model() {
+        // retrain_model now validates the model exists before forwarding
+        let model_name = "nonexistent-model";
+        let status = Status::not_found(format!(
+            "model '{}' not registered",
             model_name
         ));
-        assert_eq!(status.code(), tonic::Code::Unimplemented);
+        assert_eq!(status.code(), tonic::Code::NotFound);
         assert!(
             status.message().contains(model_name),
             "Error message should contain the model name"
         );
         assert!(
-            status.message().contains("not yet connected"),
-            "Error message should indicate the feature is not connected"
+            status.message().contains("not registered"),
+            "Error message should indicate the model is not registered"
         );
     }