Saleh Abbas-Ai/Ml & Cloud Solution Architect

• Driving digital transformation across the strategic semiconductor supply chain as part of a newly formed team building the company's AI/ML and cloud-native foundation from the ground up.
• Architected and managed a large-scale Azure Fabric & Databricks data platform supporting enterprise analytics and machine-learning workloads.
• Leading delivery of an Advanced Planning System (APS), showcasing in-house capability as a compelling alternative to off-the-shelf solutions.
• Configured MLflow tracking servers and model registries for model management, auditability, and governance-aligned artifact storage across environments.
• Implemented end-to-end CI/CD pipelines for ML workflows in Azure DevOps and defined the target MLOps architecture and enterprise migration roadmap.

Technologies used: Azure Databricks, Azure DevOps, Data Factory, Terraform, MLflow, Snowflake, Python, PySpark, Docker, Kafka

Part of setting up a newly established data department.

• Platform Design: Led the design of a new streaming and batch ingestion framework into Snowflake using Kafka Connect and Azure Data Factory.
• DevOps Enablement: Built automated CI/CD pipelines in Azure DevOps for data ingestion workflows; enforced code quality via automated linting and testing stages.
• Disaster Recovery: Implemented Hive metastore replication and failover strategies; authored DR runbooks and executed tabletop drills.

Technologies used: Snowflake, Azure Data Factory, Kafka, Terraform, Azure DevOps, Python, Power BI, Databricks, Pyspark

Group Financial Control - Data Lab

Delivered end-to-end data engineering (80–90% of each project) and machine learning for Group Financial Control, building data pipelines across Hadoop (Cloudera, Databricks), DB2/IBM Cloud, Azure, and Microsoft SQL Server and engineering hundreds of features for AI models using SQL, SAS, Python, and PySpark.

• Mortgage pricing model (Group-wide): Queried large data volumes across many servers and a data lake to build a master feature table with hundreds of columns, owning data accuracy and the selection of data points that drove a more dependable model.
• Probability-of-default model: Built a probit model in Python to predict customer default and rating classification, optimising the ROC threshold to reach 85% accuracy.
• Churn prediction: Developed a machine-learning churn model with 90% cross-validated accuracy, engineering behavioural features from log files and geodata.
• Initial-loss simulator: Created a NumPy/PySpark simulator that evaluated interest-rate campaign scenarios, visualising threshold trade-offs against the bank's initial loss in 3D with Matplotlib and Seaborn.
• Management reporting: Built a multipage interactive reporting website (Plotly, Dash) and Power BI dashboards that surfaced financial insights for senior management.
• DB2 migration: Migrated SAS and Hadoop code and tables to DB2 and authored stored procedures for recurring analytical workloads.
• Portfolio analysis: Analysed private-banking segments and profit margins, reporting directly to the CFO, and assessed Covid-19's impact across regions and industries using transaction data.
• Automation & ops: Built scheduled hourly log-file pipelines (Python, regex, PySpark) and maintained five remote Linux servers.

Technologies used: Hadoop, Hive, SAS, Spark (Scala), SQL, DB2, Python, PySpark, Scikit-learn, Statsmodels, NumPy, Pandas, Matplotlib, Seaborn, Plotly, Dash, Power BI, Azure, ETL

Built a Deep Learning Software using Multilayer perceptron to forecast the S&P 500 index. The program was built in Python using TensorFlow. The Software outperformed the Buy & Hold strategy and the ARIMA model.