LIVE WEB APP PETROPHYSICS SELF‑HOSTED
PROJECT — RESERVOIRAI

The log is wrong.
Fix it, then decide.

reservoirAI is a browser‑based workspace for repairing, analyzing, and ranking wells from digital log data. Upload LAS files, let the models repair bad‑hole sections, compare original against corrected curves, and walk out with a prioritized set of targets instead of a folder of guesses.

Open the app ↗ Read the papers

WHAT IT DOES

From LAS file to decision.

A tool built around one workflow — because the gap between a raw log and a defensible target list is where the money actually gets spent.

INGEST SHEET R-01

LAS ingestion

Upload, organize, keep per-user isolation.

Drop in LAS well logs, group them by project, and work from a single inventory. Every user’s data stays isolated while the interface stays a plain browser tab.

LASProjectsUploads
REPAIR SHEET R-02

Bad-hole repair

Machine learning where the caliper lied.

Detects bad‑hole intervals automatically, then reconstructs the affected curves with trained per‑curve models — GR, resistivity, density, neutron, and the rest — so washouts stop poisoning your averages.

scikit‑learnCurve reconstructionQC
INTERPRET SHEET R-03

Original vs. corrected

See exactly what the model changed.

Side‑by‑side log tracks make the repair auditable: original curve, reconstructed curve, and the repair mask, all on the same depth scale.

Log tracksAudit trailDepth view
PICK SHEET R-04

Tops and zones

Formation tops and pay zones, highlighted.

Mark formation tops, flag hydrocarbon zones, and generate cross‑plots for the petrophysical interpretation you’d otherwise assemble by hand.

Formation topsCross‑plotsZones
RANK SHEET R-05

Portfolio reports

Priority lists with ROI‑style metrics.

Turn per‑well analysis into a ranked program: which wells first, what the uplift looks like, and why — exported as reports you can hand to someone who wasn’t in the tool.

RankingMetricsPDF export
STACK SHEET R-06

How it runs

Flask, SQLite, and a machine that isn’t rented.

Originally deployed on EC2. Now self‑hosted behind Cloudflare Tunnel on our own hardware — same application, no monthly infrastructure bill, data that never leaves the lab.

FlaskSQLiteCloudflare Tunnel