PETRONAS AWSM — Downstream Petroleum Scheduling

PETRONAS ICT brought me in as Backend and DevOps Lead for AWSM, the downstream petroleum distribution scheduling platform. The team was strong on the operations side but needed architecture direction on AWS and Terraform. I set the technical direction, ran code reviews, and mentored engineers on serverless patterns and infrastructure as code.

The first chunk of work was getting real-time visibility into shipment status. The existing flow pulled SAP data on a schedule and dumped it into reports. I built WebSocket APIs that stream shipment events as they happen, plus Lambda functions and API Gateway endpoints feeding the operations dashboard. When a truck loads or a shipment dispatches, the console updates immediately. No more refreshing a stale report.

The second chunk was the ETL layer. Data was landing in MySQL and the warehouse from multiple sources, each with its own format and cadence. I wrote Step Functions workflows that handle the orchestration: extract, transform, load, with retries and error handling so a bad source row does not take down the whole pipeline. Once the workflows were stable, the data science team could plug their forecasting models in against a predictable schema instead of fighting input quality.

The third chunk was cost. EC2 was the bulk of the spend, and the data science workloads were hungry. I worked with the Data Science team to right-size the instance types they were using and shift the batchable jobs onto cheaper compute. The spend stayed flat as we added more dashboards and more model runs.

I also ran the stakeholder side. Presented roadmaps and progress to leadership, kept everyone aligned on what was shipping and when. Twelve-month engagement, downstream operations team that actually used what we built, and a data layer the data science team could build on instead of work around.