UDF registry file:

# Databricks notebook source
# MAGIC %pip install mlflow

# COMMAND ----------

import dlt

# COMMAND ----------

import mlflow
from pyspark.sql.functions import struct, col
logged_model = 'runs:/xxxx/model_overstort_all_devices'

# Load model as a Spark UDF. Override result_type if the model does not return double values.
loaded_model = mlflow.pyfunc.spark_udf(spark, model_uri=logged_model, result_type='double')

spark.udf.register("detect_anomaly", loaded_model)

Actual pipeline code:
CREATE OR REFRESH STREAMING LIVE TABLE iot_overstort_predict
COMMENT "Predicting anomalies in measurement data using the Isolation Forest model"
TBLPROPERTIES ("quality" = "gold")
AS SELECT 
    utc_timestamp,
    measurement_value_float,
    topic, 
    YEAR(utc_timestamp) as year, 
    MONTH(utc_timestamp) as month, 
    DAY(utc_timestamp) as day, 
    hour(utc_timestamp) as hour,
    detect_anomaly(measurement_value_float)
FROM STREAM(aqf_tda.test.xxx)

Actual pipeline: