nextreservoir.com · Subsurface AI
The work engine
for the individual
subsurface engineer.
Ask it a question and it doesn't just answer — it runs the calculation, writes the code, builds the plot, and cites every source down to the page. All in the browser. No install.
Example session · Production Dashboard
Build a production dashboard for Well P-1 from the monthly allocation export — oil, gas and water rates, water cut, and cumulative oil.
P-1 is producing 2,140 STB/d of oil at a 38 % water cut P1_allocation_2025.xlsx · sheet “monthly” · Session upload, with cumulative oil of 1.84 MMSTB to date Field_production_history.csv · Space docs. Water cut has climbed 11 points over the last six months — worth flagging before the next forecast. Dashboard below.
import pandas as pd
import plotly.graph_objects as go
# Monthly allocation export (P1_allocation_2025.xlsx)
df = pd.read_excel("P1_allocation_2025.xlsx", sheet_name="monthly")
df["water_cut"] = df["water_bbl"] / (df["oil_bbl"] + df["water_bbl"])
df["cum_oil"] = df["oil_bbl"].cumsum() / 1e6 # MMSTB
fig = go.Figure()
fig.add_scatter(x=df["date"], y=df["oil_rate"], name="Oil", line_color="#2B6C2B")
fig.add_scatter(x=df["date"], y=df["water_rate"], name="Water", line_color="#1657A3")
print(f"Latest oil rate = {df['oil_rate'].iloc[-1]:,.0f} STB/d") # → 2,140The workspace
This is what the work looks like —
end to end, in one screen.
The mismatch exceeds the recommended QC tolerance of ±10 % SPE-123456 · p.8. Recommend updating the SGOF table before the next history-match cycle.
Core capabilities
It doesn't answer questions.
It completes the work.
Every interaction produces something you can defend: a cited source, inspectable code, a reproducible plot, or a draft deliverable. Not a chatbot output.
Every claim traces to a source
Inline citation pills show the document, page, and knowledge tier for every factual statement. A structural validation layer runs on every answer — hallucinated citations are stripped automatically before you see them.
The AI never does arithmetic in its head
All calculations are written and executed as Python. The code is always visible and downloadable. You run the same script in your own environment and get the same number — that is the reliability guarantee.
Every major simulator — side by side
Next Reservoir generates keyword blocks and Python setups for the simulator you use. You run them in your licensed environment and paste or upload the result back. No vendor lock-in.
QC packs, plots, and report drafts
Upload your history-match output, SCAL tables, or well logs and get a structured QC pack — flagged anomalies, standard cross-plots in house style, and a draft technical note, ready to defend in peer review.
Open a browser. That's it.
No software to install, no IT approval, no VPN. The full capability — calculations, code, corpus search — runs in the cloud. The copy-down path lets you take generated code to your local licensed environment whenever needed.
Your data is yours
What you upload stays in your private workspace and is never pooled into a shared corpus. EU-resident inference. GDPR-compliant architecture. Row-level security enforces tenant isolation at the database level — not by convention.
Citation pills
The trust mechanism —
auditable by design.
In a domain where a wrong number is expensive, auditability is not a nice-to-have. It is the product. Every citation tells you exactly where the claim came from.
Source document & page
Kazemi et al. · SPEJ 1976 · p.4
Knowledge tier
Admin library Space docs Your Boxed library Session upload
Validation guarantee
Every cited chunk verified to be in the retrieved set. Hallucinated citations are stripped and flagged — automatically, on every answer.
How it works
Four steps from
question to deliverable.
The pipeline is visible at every stage — you can inspect what was retrieved, what code was run, and what the source actually says.
Ask or upload
Type a question, paste a dataset, or upload a PDF, LAS, SCAL table, or simulation output. The system detects the format and renders a default view immediately.
Retrieve & rank
Hybrid search across all four knowledge tiers — semantic and keyword combined — retrieves the best matching chunks, re-ranked by a cross-encoder reranker for accuracy.
Compute & cite
The AI writes Python, runs it in a sandboxed environment, and grounds every claim in the retrieved sources. Citations are validated structurally before you see them.
Deliver
A cited answer, a Plotly chart in house style, downloadable code, or a QC pack draft — ready to defend, share, or compile into a report.
Visualization
See your subsurface data —
the moment you upload it.
Structure maps, well logs and LAS curves, SCAL and PVT tables, grid and simulation output, production histories — Next Reservoir detects the format and renders a clean, house-style view on upload. The full range of subsurface data formats, visualized in the browser without a single line of setup.
Simulator support
Work in your simulator.
Not ours.
Next Reservoir generates and QC's the inputs. You run them in your licensed environment. Results come back for analysis and reporting.
| Simulator | Keyword generation | Results analysis | How it runs |
|---|---|---|---|
| OPM Flow | ✓ Full keyword blocks | ✓ SMSPEC, EGRID | Cloud sandbox — full loop in-browser |
| Commercial simulators | ✓ Full keyword decks | ✓ resdata / output upload | Copy-down path to your licensed install |
Knowledge library
Your knowledge and the field's knowledge —
in one ranked hierarchy.
Every answer searches across four tiers simultaneously. Citations show which tier a source belongs to. Your data always outranks the shared corpus.
Session — your current upload
Files you drag in during this working session. Highest retrieval priority. Expires after 60 days; metadata preserved for citation integrity.
Space — your project library
Documents shared across all sessions in a project. Your field data, reports, and references — searchable throughout the engagement.
Boxed Library — your personal corpus
Portable — belongs to your account, not an employer. Papers, methods, references you've curated over a career. Follows you across every project.
Admin Library — the shared foundation
Curated technical corpus: SPE papers, SCAL & PVT literature, simulator manuals, open data. Human-vetted before it goes live. Available to all subscribers.
“A general AI tool can summarise a paper. It cannot tell you that your Krg endpoint is 18 % below the simulator default — and show you the code that proves it.”Built for the subsurface · Nothing horizontal comes close
House style
Plots that look like your work,
not the AI's defaults.
Every generated plot uses the Next Reservoir fluid colour system — the conventions your reviewers and clients already expect. User-overridable per project.
Line conventions: solid = observed · – – – dashed = simulated · · · · dotted = forecast. Baked into every Plotly output.
Get started
The work is waiting.
Start the session.
Next Reservoir is in closed early access for individual engineers and consultants. No install. No team required. One subscription, your entire subsurface workflow.