Your company knows
more than it
can find.

AI knowledge retrieval for geoscience, subsurface, and technical energy teams.

Kataba turns legacy reports, well files, maps, scanned documents, and interpretations into source-cited answers your team can verify.

See the Demo Discuss a Pilot
Geological cross-section showing subsurface layers and a drilling rig

The knowledge exists. Finding it is the problem.

Subsurface teams often have decades of valuable technical knowledge scattered across PDFs, scanned reports, PowerPoints, maps, well files, shared drives, inboxes, and retired experts' laptops. When that knowledge is hard to retrieve, teams lose time, recreate prior work, and make decisions without full context.

01 Reports no one can search Technical PDFs, scans, maps, and presentations scattered across drives and inboxes
02 Knowledge that leaves with people Years of asset intuition disappear when senior experts retire or move on
03 Work that gets done twice Teams rerun studies because prior analyses are buried or unknown
04 Decisions that wait on searches Asset reviews slow down while teams hunt for the right report, map, or interpretation

The cost of buried knowledge is well documented

20%

of the workweek knowledge workers spend searching for and gathering information

McKinsey Global Institute

75%

reduction in engineering-modification update time after deploying a knowledge-management system

Schlumberger · Wharton / Mack Institute

$650M

in total savings attributed to knowledge-management programs over eight years

Chevron · Wharton / Mack Institute

70%

lower search time across roughly two million documents with AI document retrieval

AWS · DataStax · Shorthills AI case study

Kataba applies this same economic logic to subsurface teams: reduce search time, prevent duplicated technical work, preserve institutional memory, and make decisions with cited evidence instead of buried knowledge.

The Solution

Ask your technical documents a question. Get an answer with evidence.

Kataba retrieves relevant passages from your document set, ranks the evidence, and generates a concise answer with citations back to the original source.

Document retrieval from subsurface knowledge layers
01

Designed for private deployment

Kataba can be deployed in a controlled environment so proprietary documents remain isolated and traceable.

02

Plain-English questions

Ask by basin, field, formation, well, interval, or concept — the way you'd ask a colleague. No query language. No training required.

03

Every answer is cited

Source document, page number, and retrieved evidence are shown so users can verify before acting.

kataba · demo corpus · 127 documents indexed
RD
What evidence do we have for source rock maturity in the Cretaceous interval? Summarize our internal studies from the last 10 years.
K
In this demo corpus, several indexed documents discuss high thermal maturity (Ro 1.1–1.4%) between 3,800–4,200 m depth. 📄 Robertson_2019 p.14 A later basin model raised maturity window estimates by ~8% after revised heat-flow calibration. 📄 BasinModel_2022 p.7 One dissenting memo flags a local depocenter anomaly in the northern block. 📄 NBlock_Memo_2021 p.3
Which formations are candidates for secondary migration pathways?
How We Work

Pilot with a focused document set.

We start with a limited, well-defined corpus so retrieval quality, citation accuracy, and high-value workflows can be validated before any broader deployment.

Typical corpus

  • Basin reports and field studies
  • Well files and completion reports
  • Scanned PDFs and legacy maps
  • Technical presentations and memos
  • Interpretations and internal analyses

Typical pilot goals

  • Test retrieval quality and citation accuracy
  • Identify high-value workflows
  • Estimate time saved per team member
  • Validate knowledge coverage
  • Define deployment requirements
Discuss a 4–6 Week Pilot

Built by geoscience and software operators

Kataba is a US-based team. Two of us hold PhDs in geosciences, one is actively working in the oil industry, and two previously worked together on applied machine learning at Apple. We are speaking with early teams who want to validate the value before committing to a broader deployment.

From katabasis — descent into the earth

Named for the descent.

Kataba takes its name from katabasis (κατάβασις) — the ancient Greek word for a journey downward. Like the geoscientist's path into the subsurface, Kataba uncovers what's buried beneath the surface, turning complexity into understanding.

Descend. Discover. Decide.

Want to test Kataba on a
real document set?

We are speaking with geoscience and subsurface teams dealing with legacy reports, maps, scans, and fragmented technical knowledge.

Prefer email? Reach us at hello@kataba.ai