- Architected and deployed an enterprise-grade RAG platform summarizing ~10 LNG
docs/day (~30 pages each), eliminating manual review by a 10-engineer team
- Designed scalable ETL pipelines in Python and Databricks processing ~500 MB of operational
data daily in near real-time
- Integrated custom embeddings, vectorization, high-performance indexing, and an LLM-driven chat
interface
- Adopted by ~50 traders and analysts across Petronas; led and mentored 3 new
engineers
PythonPySparkDelta LakeDatabricksMedallion
Arch.Azure DevOpsRAGNLPVector DB