I build machine learning systems and agentic AI workflows for upstream and midstream operators — the kind that survive contact with messy SCADA exports, legacy WITSML feeds, and the engineer who insists everything still lives in a 40-tab Excel file.
Production pipelines for forecasting, anomaly detection, and physics-informed models. Built to run on the rig, not just in the notebook.
LLM agents that read daily drilling reports, draft DOIs, and surface non-productive time root causes — with humans firmly in the loop.
Refactoring decade-old reservoir & commercial models into validated, version-controlled tools — without forcing your team off Excel.
Power BI, Streamlit, and custom web dashboards wired into PI, OSDU, and SAP — built so a field engineer trusts the number on screen.
Sit with the team that owns the decision. Understand the operational pain — not the algorithm someone read about on LinkedIn.
Inventory the data: SCADA, PI, OSDU, drilling reports, the Excel that lives on someone's desktop. Surface what's signal, what's swamp.
Ship a working model or agent in weeks, not quarters. Tested against historical events you actually remember.
Wire it into a dashboard, an alerting system, or an Excel-friendly API. Versioning, monitoring, evals — the boring parts.
Document, train your team, and leave. I'm independent — my goal is to make myself replaceable, not embedded forever.
Trained a sequence model on 6 years of mud-log and rig-state data; deployed as a real-time advisor inside the existing rig dashboard.
LLM agents that pull from SAP, allocation systems, and email attachments to draft a board-ready morning brief — reviewed by a human in 10 minutes.
Per-well choke-setting recommendations from a physics-informed ML model, surfaced in a Streamlit tool the production engineers actually open every shift.
Eight years across two majors and one independent — first as a reservoir engineer in Jakarta and Houston, then leading the data science function for an upstream operator's digital transformation.
RFour Energy Corp. was founded in 2023 because the most interesting ML problems in oil & gas don't sit inside one company — they sit at the seam between the rig, the spreadsheet, and the boardroom. That's where we work.