Worked on enhancing artifact retrieval compatibility for Databricks Serverless environments within the mlflow/mlflow repository. Addressed artifact visibility issues by modifying the spark_udf implementation to exclude the use_dbconnect_artifact path, and introduced an environment variable to control compatibility behavior across deployments. Enabled UDF executors to fetch models directly from the artifact store when spark.addArtifact could not surface artifacts, improving reliability of model loading in serverless contexts. The work leveraged Python, Databricks, and MLflow, focusing on cloud computing and data engineering principles. These changes aligned with Databricks Serverless requirements and contributed to more robust cross-environment deployment workflows.