diff --git a/gateway/app/routes/training.py b/gateway/app/routes/training.py
index fbad9318..0bea905f 100644
--- a/gateway/app/routes/training.py
+++ b/gateway/app/routes/training.py
@@ -13,38 +13,6 @@ from app.core.config import settings
 logger = logging.getLogger(__name__)
 router = APIRouter()
 
-@router.post("/train")
-async def start_training(request: Request):
-    """Proxy training request to training service"""
-    try:
-        body = await request.body()
-        auth_header = request.headers.get("Authorization")
-        
-        async with httpx.AsyncClient(timeout=30.0) as client:
-            response = await client.post(
-                f"{settings.TRAINING_SERVICE_URL}/train",
-                content=body,
-                headers={
-                    "Content-Type": "application/json",
-                    "Authorization": auth_header
-                }
-            )
-            
-            return JSONResponse(
-                status_code=response.status_code,
-                content=response.json()
-            )
-            
-    except httpx.RequestError as e:
-        logger.error(f"Training service unavailable: {e}")
-        raise HTTPException(
-            status_code=503,
-            detail="Training service unavailable"
-        )
-    except Exception as e:
-        logger.error(f"Training error: {e}")
-        raise HTTPException(status_code=500, detail="Internal server error")
-
 @router.get("/status/{training_job_id}")
 async def get_training_status(training_job_id: str, request: Request):
     """Get training job status"""
@@ -130,3 +98,35 @@ async def get_training_jobs(
     except Exception as e:
         logger.error(f"Get training jobs error: {e}")
         raise HTTPException(status_code=500, detail="Internal server error")
+
+@router.post("/jobs")
+async def start_training_job(request: Request):
+    """Start a new training job - Proxy to training service"""
+    try:
+        body = await request.body()
+        auth_header = request.headers.get("Authorization")
+        
+        async with httpx.AsyncClient(timeout=30.0) as client:
+            response = await client.post(
+                f"{settings.TRAINING_SERVICE_URL}/training/jobs",  # Correct path
+                content=body,
+                headers={
+                    "Content-Type": "application/json",
+                    "Authorization": auth_header
+                }
+            )
+            
+            return JSONResponse(
+                status_code=response.status_code,
+                content=response.json()
+            )
+            
+    except httpx.RequestError as e:
+        logger.error(f"Training service unavailable: {e}")
+        raise HTTPException(
+            status_code=503,
+            detail="Training service unavailable"
+        )
+    except Exception as e:
+        logger.error(f"Start training job error: {e}")
+        raise HTTPException(status_code=500, detail="Internal server error")
\ No newline at end of file
diff --git a/services/data/app/external/aemet.py b/services/data/app/external/aemet.py
index 66c4ffe3..bf1242eb 100644
--- a/services/data/app/external/aemet.py
+++ b/services/data/app/external/aemet.py
@@ -136,9 +136,9 @@ class AEMETClient(BaseAPIClient):
         try:
             # Madrid area stations (simplified)
             madrid_stations = {
-                "3195": {"lat": 40.4168, "lon": -3.7038, "name": "Madrid Centro"},
-                "3196": {"lat": 40.4518, "lon": -3.7246, "name": "Madrid Norte"},
-                "3197": {"lat": 40.3833, "lon": -3.7167, "name": "Madrid Sur"}
+                "3195": {"lat": 40.4117, "lon": -3.6780, "name": "Madrid Centro"},
+                "3129": {"lat": 40.4677, "lon": -3.5552, "name": "Madrid Norte"}, 
+                "3197": {"lat": 40.2987, "lon": -3.7216, "name": "Madrid Sur"} 
             }
             
             closest_station = None
diff --git a/services/data/app/external/madrid_opendata.py b/services/data/app/external/madrid_opendata.py
index d0091ab7..2bd655c0 100644
--- a/services/data/app/external/madrid_opendata.py
+++ b/services/data/app/external/madrid_opendata.py
@@ -630,10 +630,386 @@ class MadridOpenDataClient(BaseAPIClient):
             "source": "synthetic"
         }
     
-    # Placeholder methods for completeness
     async def get_historical_traffic(self, latitude: float, longitude: float, start_date: datetime, end_date: datetime) -> List[Dict[str, Any]]:
-        """Get historical traffic data"""
-        return []
+        """Get historical traffic data from Madrid Open Data
+        
+        Args:
+            latitude: Location latitude
+            longitude: Location longitude  
+            start_date: Start date for historical data
+            end_date: End date for historical data
+            
+        Returns:
+            List of historical traffic data dictionaries
+        """
+        try:
+            logger.debug("Fetching Madrid historical traffic data", 
+                        lat=latitude, lon=longitude, 
+                        start=start_date, end=end_date)
+            
+            historical_data = []
+            
+            # Generate historical data using synthetic generation for periods before API availability
+            # or when real data is not available
+            if (end_date - start_date).days <= 90:  # Reasonable range for synthetic data
+                historical_data = await self._generate_historical_traffic(latitude, longitude, start_date, end_date)
+                logger.info("Generated synthetic historical traffic data", 
+                           records=len(historical_data))
+            else:
+                logger.warning("Date range too large for historical traffic data", 
+                             days=(end_date - start_date).days)
+                return []
+            
+            # Try to fetch real data if API key is available and for recent dates
+            if hasattr(self, 'api_key') and self.api_key:
+                try:
+                    real_data = await self._fetch_real_historical_traffic(latitude, longitude, start_date, end_date)
+                    if real_data:
+                        # Merge real data with synthetic data or replace synthetic data
+                        historical_data = real_data
+                        logger.info("Fetched real historical traffic data", 
+                                   records=len(real_data))
+                except Exception as e:
+                    logger.warning("Failed to fetch real historical data, using synthetic", error=str(e))
+            
+            return historical_data
+            
+        except Exception as e:
+            logger.error("Error getting historical traffic data", error=str(e))
+            return []
+    
+    async def _fetch_real_historical_traffic(self, latitude: float, longitude: float, start_date: datetime, end_date: datetime) -> List[Dict[str, Any]]:
+        """Fetch real historical traffic data from Madrid Open Data portal
+        
+        Madrid provides historical CSV files by month at:
+        https://datos.madrid.es/egob/catalogo/[ID]-[YEAR]-[MONTH]-trafico-historico.csv
+        """
+        try:
+            historical_data = []
+            current_date = start_date.replace(day=1)  # Start from beginning of month
+            
+            while current_date <= end_date:
+                try:
+                    # Madrid historical traffic CSV URL pattern
+                    year = current_date.year
+                    month = current_date.month
+                    
+                    # Try different URL patterns based on Madrid Open Data structure
+                    historical_urls = [
+                        f"https://datos.madrid.es/egob/catalogo/300217-{year}-{month:02d}-trafico-historico.csv",
+                        f"https://datos.madrid.es/egob/catalogo/trafico-historico-{year}-{month:02d}.csv",
+                        f"https://datos.madrid.es/egob/catalogo/{year}{month:02d}-trafico-historico.csv"
+                    ]
+                    
+                    for url in historical_urls:
+                        csv_data = await self._fetch_historical_csv(url)
+                        if csv_data:
+                            # Parse CSV and filter by location
+                            month_data = await self._parse_historical_csv(csv_data, latitude, longitude, start_date, end_date)
+                            historical_data.extend(month_data)
+                            logger.debug("Fetched historical data for month", 
+                                        year=year, month=month, records=len(month_data))
+                            break
+                    
+                    # Move to next month
+                    if current_date.month == 12:
+                        current_date = current_date.replace(year=current_date.year + 1, month=1)
+                    else:
+                        current_date = current_date.replace(month=current_date.month + 1)
+                        
+                except Exception as e:
+                    logger.warning("Error fetching data for month", 
+                                 year=current_date.year, month=current_date.month, error=str(e))
+                    # Move to next month even on error
+                    if current_date.month == 12:
+                        current_date = current_date.replace(year=current_date.year + 1, month=1)
+                    else:
+                        current_date = current_date.replace(month=current_date.month + 1)
+            
+            return historical_data
+            
+        except Exception as e:
+            logger.error("Error fetching real historical traffic data", error=str(e))
+            return []
+    
+    async def _fetch_historical_csv(self, url: str) -> Optional[str]:
+        """Fetch historical CSV data from Madrid Open Data"""
+        try:
+            import httpx
+            
+            headers = {
+                'User-Agent': 'Mozilla/5.0 (compatible; Madrid-Traffic-Client/1.0)',
+                'Accept': 'text/csv,application/csv,text/plain,*/*',
+                'Accept-Language': 'es-ES,es;q=0.9,en;q=0.8',
+            }
+            
+            async with httpx.AsyncClient(timeout=60.0, headers=headers) as client:
+                logger.debug("Fetching historical CSV", url=url)
+                response = await client.get(url)
+                
+                if response.status_code == 200:
+                    content = response.text
+                    if content and len(content) > 100:  # Ensure we got actual data
+                        logger.debug("Successfully fetched CSV", 
+                                   url=url, size=len(content))
+                        return content
+                else:
+                    logger.debug("CSV not found", url=url, status=response.status_code)
+                    
+        except Exception as e:
+            logger.debug("Error fetching CSV", url=url, error=str(e))
+            
+        return None
+    
+    async def _parse_historical_csv(self, csv_content: str, latitude: float, longitude: float, start_date: datetime, end_date: datetime) -> List[Dict[str, Any]]:
+        """Parse Madrid historical traffic CSV and filter by location and date range"""
+        try:
+            import csv
+            from io import StringIO
+            
+            historical_records = []
+            csv_reader = csv.DictReader(StringIO(csv_content), delimiter=';')
+            
+            # Get the nearest measurement points to our coordinates
+            measurement_points = await self._get_measurement_points_near_location(latitude, longitude)
+            target_point_ids = [point['id'] for point in measurement_points[:3]]  # Use 3 nearest points
+            
+            for row in csv_reader:
+                try:
+                    # Parse Madrid CSV format
+                    # Expected columns: fecha, hora, idelem, intensidad, ocupacion, carga, nivelServicio, etc.
+                    
+                    # Extract date and time
+                    if 'fecha' in row and 'hora' in row:
+                        date_str = row.get('fecha', '').strip()
+                        time_str = row.get('hora', '').strip()
+                        
+                        # Parse Madrid date format (usually DD/MM/YYYY)
+                        if date_str and time_str:
+                            try:
+                                # Try different date formats
+                                for date_format in ['%d/%m/%Y', '%Y-%m-%d', '%d-%m-%Y']:
+                                    try:
+                                        record_date = datetime.strptime(f"{date_str} {time_str}", f"{date_format} %H:%M")
+                                        break
+                                    except ValueError:
+                                        continue
+                                else:
+                                    continue  # Skip if no date format worked
+                                
+                                # Check if record is in our date range
+                                if not (start_date <= record_date <= end_date):
+                                    continue
+                                    
+                            except ValueError:
+                                continue
+                    else:
+                        continue
+                    
+                    # Check if this record is from a measurement point near our location
+                    point_id = row.get('idelem', '').strip()
+                    if point_id not in target_point_ids:
+                        continue
+                    
+                    # Parse traffic data
+                    traffic_record = {
+                        "date": record_date,
+                        "traffic_volume": self._safe_int(row.get('intensidad', '0')),
+                        "occupation_percentage": self._safe_int(row.get('ocupacion', '0')),
+                        "load_percentage": self._safe_int(row.get('carga', '0')),
+                        "service_level": self._safe_int(row.get('nivelServicio', '0')),
+                        "measurement_point_id": point_id,
+                        "measurement_point_name": row.get('descripcion', f'Point {point_id}'),
+                        "road_type": row.get('tipo_elem', 'URB'),
+                        "source": "madrid_opendata_historical"
+                    }
+                    
+                    # Calculate derived metrics
+                    service_level = traffic_record['service_level']
+                    if service_level == 0:  # Fluid
+                        congestion_level = "low"
+                        avg_speed = 45
+                        pedestrian_multiplier = 1.0
+                    elif service_level == 1:  # Dense
+                        congestion_level = "medium"
+                        avg_speed = 25
+                        pedestrian_multiplier = 1.5
+                    elif service_level == 2:  # Congested
+                        congestion_level = "high"
+                        avg_speed = 15
+                        pedestrian_multiplier = 2.0
+                    else:  # Cut/Blocked
+                        congestion_level = "blocked"
+                        avg_speed = 5
+                        pedestrian_multiplier = 0.5
+                    
+                    traffic_record.update({
+                        "congestion_level": congestion_level,
+                        "average_speed": avg_speed,
+                        "pedestrian_count": int(100 * pedestrian_multiplier)
+                    })
+                    
+                    historical_records.append(traffic_record)
+                    
+                except Exception as e:
+                    logger.debug("Error parsing CSV row", error=str(e))
+                    continue
+            
+            return historical_records
+            
+        except Exception as e:
+            logger.error("Error parsing historical CSV", error=str(e))
+            return []
+    
+    async def _get_measurement_points_near_location(self, latitude: float, longitude: float) -> List[Dict[str, Any]]:
+        """Get measurement points near the specified location"""
+        try:
+            # Try to fetch current traffic data to get measurement points
+            current_traffic = await self._fetch_traffic_xml_data(self.traffic_endpoints[0])
+            
+            if current_traffic:
+                # Calculate distances and sort by proximity
+                points_with_distance = []
+                for point in current_traffic:
+                    if point.get('latitude') and point.get('longitude'):
+                        distance = self._calculate_distance(
+                            latitude, longitude,
+                            point['latitude'], point['longitude']
+                        )
+                        points_with_distance.append({
+                            'id': point.get('idelem'),
+                            'distance': distance,
+                            'latitude': point['latitude'],
+                            'longitude': point['longitude'],
+                            'name': point.get('descripcion', '')
+                        })
+                
+                # Sort by distance and return closest points
+                points_with_distance.sort(key=lambda x: x['distance'])
+                return points_with_distance[:5]  # Return 5 closest points
+            
+            # Fallback: return synthetic point IDs based on Madrid geography
+            return [
+                {'id': 'madrid_centro_01', 'distance': 1.0},
+                {'id': 'madrid_centro_02', 'distance': 2.0},
+                {'id': 'madrid_centro_03', 'distance': 3.0}
+            ]
+            
+        except Exception as e:
+            logger.warning("Error getting measurement points", error=str(e))
+            return [{'id': 'madrid_default', 'distance': 0.0}]
+    
+    async def _generate_historical_traffic(self, latitude: float, longitude: float, start_date: datetime, end_date: datetime) -> List[Dict[str, Any]]:
+        """Generate synthetic historical traffic data for the specified period
+        
+        This method creates realistic historical traffic patterns based on:
+        - Time of day patterns
+        - Day of week patterns  
+        - Seasonal variations
+        - Random variations for realism
+        """
+        try:
+            import random
+            from datetime import timedelta
+            
+            historical_data = []
+            current_date = start_date
+            
+            # Seed random for consistent but varied data
+            random.seed(hash(f"{latitude}{longitude}"))
+            
+            while current_date <= end_date:
+                # Generate 24 hourly records for each day
+                for hour in range(24):
+                    record_time = current_date.replace(hour=hour, minute=0, second=0, microsecond=0)
+                    
+                    # Base traffic calculation
+                    base_traffic = 100
+                    hour_of_day = record_time.hour
+                    day_of_week = record_time.weekday()  # 0=Monday, 6=Sunday
+                    month = record_time.month
+                    
+                    # Time of day patterns
+                    if 7 <= hour_of_day <= 9:  # Morning rush
+                        traffic_multiplier = 2.2 + random.uniform(-0.3, 0.3)
+                        congestion = "high"
+                        avg_speed = 15 + random.randint(-5, 5)
+                    elif 18 <= hour_of_day <= 20:  # Evening rush
+                        traffic_multiplier = 2.5 + random.uniform(-0.4, 0.4)
+                        congestion = "high"
+                        avg_speed = 12 + random.randint(-3, 8)
+                    elif 12 <= hour_of_day <= 14:  # Lunch time
+                        traffic_multiplier = 1.6 + random.uniform(-0.2, 0.2)
+                        congestion = "medium"
+                        avg_speed = 25 + random.randint(-5, 10)
+                    elif 22 <= hour_of_day or hour_of_day <= 6:  # Night
+                        traffic_multiplier = 0.3 + random.uniform(-0.1, 0.2)
+                        congestion = "low"
+                        avg_speed = 50 + random.randint(-10, 15)
+                    else:  # Regular hours
+                        traffic_multiplier = 1.0 + random.uniform(-0.2, 0.2)
+                        congestion = "medium"
+                        avg_speed = 35 + random.randint(-10, 10)
+                    
+                    # Weekend adjustments
+                    if day_of_week >= 5:  # Weekend
+                        if hour_of_day in [11, 12, 13, 14, 15]:  # Weekend afternoon peak
+                            traffic_multiplier *= 1.4
+                            congestion = "medium"
+                        else:
+                            traffic_multiplier *= 0.7
+                            if congestion == "high":
+                                congestion = "medium"
+                    
+                    # Seasonal adjustments
+                    if month in [7, 8]:  # Summer - less traffic due to vacations
+                        traffic_multiplier *= 0.8
+                    elif month in [11, 12]:  # Holiday season - more traffic
+                        traffic_multiplier *= 1.1
+                    
+                    # Calculate final values
+                    traffic_volume = max(10, int(base_traffic * traffic_multiplier))
+                    avg_speed = max(10, min(60, avg_speed))
+                    
+                    # Pedestrian calculation
+                    pedestrian_base = 150
+                    if 13 <= hour_of_day <= 15:  # Lunch time
+                        pedestrian_count = int(pedestrian_base * 2.5 * random.uniform(0.8, 1.2))
+                    elif 8 <= hour_of_day <= 9 or 18 <= hour_of_day <= 20:  # Rush hours
+                        pedestrian_count = int(pedestrian_base * 2.0 * random.uniform(0.8, 1.2))
+                    else:
+                        pedestrian_count = int(pedestrian_base * 1.0 * random.uniform(0.5, 1.5))
+                    
+                    # Create traffic record
+                    traffic_record = {
+                        "date": record_time,
+                        "traffic_volume": traffic_volume,
+                        "pedestrian_count": pedestrian_count,
+                        "congestion_level": congestion,
+                        "average_speed": avg_speed,
+                        "occupation_percentage": min(100, traffic_volume // 2),
+                        "load_percentage": min(100, traffic_volume // 3),
+                        "measurement_point_id": f"madrid_historical_{hash(f'{latitude}{longitude}') % 1000}",
+                        "measurement_point_name": f"Madrid Historical Point ({latitude:.4f}, {longitude:.4f})",
+                        "road_type": "URB",
+                        "source": "synthetic_historical"
+                    }
+                    
+                    historical_data.append(traffic_record)
+                
+                # Move to next day
+                current_date += timedelta(days=1)
+            
+            logger.info("Generated historical traffic data", 
+                       records=len(historical_data),
+                       start=start_date,
+                       end=end_date)
+            
+            return historical_data
+            
+        except Exception as e:
+            logger.error("Error generating historical traffic data", error=str(e))
+            return []
     
     async def get_events(self, latitude: float, longitude: float, radius_km: float = 5.0) -> List[Dict[str, Any]]:
         """Get traffic incidents and events"""
diff --git a/services/data/app/services/traffic_service.py b/services/data/app/services/traffic_service.py
index 3bf815dd..1735911d 100644
--- a/services/data/app/services/traffic_service.py
+++ b/services/data/app/services/traffic_service.py
@@ -90,8 +90,40 @@ class TrafficService:
                     average_speed=record.average_speed,
                     source=record.source
                 ) for record in db_records]
+            
+            # If not in database, fetch from API and store
+            logger.debug("Fetching historical data from MADRID OPEN DATA")
+            traffic_data = await self.madrid_client.get_historical_traffic(
+                latitude, longitude, start_date, end_date
+            )
+            
+            if traffic_data:
+                # Store in database for future use
+                try:
+                    for data in traffic_data:
+                        traffic_record = TrafficData(
+                            id = id,
+                            location_id = location_id,
+                            date = data.get('date', datetime.now()),
+                            traffic_volume = data.get('traffic_volume'),
+                            pedestrian_count = data.get('pedestrian_count'),
+                            congestion_level = data.get('congestion_level'),
+                            average_speed = data.get('average_speed'),
+                            source = "Madrid Open Data",
+                            raw_data = str(data),
+                            created_at = data.get('created_at'),
+                        )
+                        db.add(traffic_record)
+                    
+                    await db.commit()
+                    logger.debug("Historical data stored in database", count=len(traffic_record))
+                except Exception as db_error:
+                    logger.warning("Failed to store historical data in database", error=str(db_error))
+                    await db.rollback()
+                
+                return [TrafficDataResponse(**item) for item in traffic_record]
             else:
-                logger.debug("No historical traffic data found in database")
+                logger.warning("No historical traffic data received")
                 return []
                 
         except Exception as e:
diff --git a/services/training/app/main.py b/services/training/app/main.py
index edc1e377..d83f526d 100644
--- a/services/training/app/main.py
+++ b/services/training/app/main.py
@@ -177,7 +177,6 @@ async def global_exception_handler(request: Request, exc: Exception):
 # Authentication is handled by API Gateway
 app.include_router(
     training.router,
-    prefix="/training",
     tags=["training"]
 )