Mohamed Achraf Baiz

Developed a Python REST API using FastAPI packaged within an Azure Function, containerized it as a Docker image, and deployed it on Azure Function App for scalable serverless execution.

Built a LangChain / LangGraph–based multi-agent system (3-agent graph) leveraging OpenAI to perform Text-to-SQL generation, including SQL creation via Databricks Genie.

Integrated SQLGlot for sql parsing, validation, correction, and adaptation, enabling automated remediation of malformed queries before execution.

Implemented logic to execute validated SQL against Databricks Genie API or custom queries on Databricks SQL Warehouse, followed by a downstream agent to interpret query results against the appropriate table/view.

Built a Databricks Bundle pipeline to register a custom embedding model into MLflow Unity Catalog, and automated model serving by creating/updating serving endpoints.

Created an additional Databricks Vector Search endpoint to build and maintain vector indexes using delta_sync on target Delta tables.

Implemented data integration pipelines on Databricks 15.4 LTS using PySpark 3.5.5, automating ingestion, transformation, and optimization workflows across Delta Lake storage layers (bronze, silver, gold).

Engineered a custom data catalog to track pipeline lineage, schema evolution, and transformation logic, enabling transparent governance and auditability across the multi-layer Delta architecture.

Built scalable, fault-tolerant ETL/ELT processes leveraging Delta tables, optimized cluster configurations, and Databricks-native features (e.g., Auto Loader, Delta Live Tables patterns) to ensure data quality, consistency, and high performance.

Designed and deployed Azure Machine Learning pipelines and Streamlit applications, enabling scalable, reproducible model deployments and automated workflows via Azure Container Registry.

Developed a FastAPI/Azure Functions backend interfacing with MATLAB Production Server, supporting simulations, API management, OpenAPI documentation, and ADLS Gen2 integration.

Built high-performance Python daemons and weather-driven data pipelines for industrial system integration, real-time telemetry ingestion into InfluxDB, and DER forecasting.

Created dashboards with Chronograf, implemented structured logging (Loguru), and enforced rigorous data validation with Pydantic and Pandera.

Automated testing, CI/CD (GitHub Actions), dependency management (JFrog Artifactory), and container security (Trivy, JFrog Xray) to ensure quality, reproducibility, and secure deployments.

Engineered modular, configuration-driven architectures (YAML + Python validation), optimized columnar data handling with PyArrow, and maintained documentation and changelogs using Sphinx and git-cliff.

Developed Python-based ETL pipelines to ingest, integrate, and process data from Salesforce, Google Ads API, Adobe Omniture API 1.4, and Oracle Siebel CRM (PyODBC), handling schema normalization, deduplication, historical tracking, and high-volume datasets for marketing and sales analytics. Implemented CI/CD pipelines with AWS CodeCommit, CodeBuild, Lambda, ECR, and CodeArtifact, containerized ETL services with Docker, and orchestrated batch workflows using Airflow for automated, reliable, and reproducible deployments. Leveraged AWS Glue Crawlers and Data Catalog for metadata management, Amazon Athena for querying and aggregation, AWS RDS (PostgreSQL) for transactional support, and AWS DMS for daily data migrations across heterogeneous sources. Optimized AWS S3 data lake storage with partitioning, lifecycle policies, and integration with downstream analytics; automated operations with Python and Shell scripts, and monitored pipelines with dashboards (Chronograf). Applied domain expertise in automotive marketing and sales, collaborating with cross-functional teams to validate data quality, align pipelines with business KPIs, and ensure actionable insights for reporting and analytics.