AI Systems Index

What we work with.

Seven areas. Choose one to see what it is, what we build and where it applies.

Pattern reference

When each pattern fits.

Explanatory material, not a list of delivered work.

Retrieval-augmented generation answers a question from retrieved source passages instead of from a model’s memory alone.

Fits: When answers must come from a specific body of documents and be checkable.
Needs: Well-processed documents, good retrieval, and citations back to the source.

An agent uses a model to decide the next step in a task and calls tools to carry it out.

Fits: When work has several steps and the path depends on what is found along the way.
Needs: Narrow, well-defined tools, logging, and review points for decisions that matter.

Turning documents into structured, searchable content: text, layout, tables and meaning.

Fits: When information is locked in files that people currently read by hand.
Needs: Reliable parsing for the document types involved, and a check on extraction quality.

Software that detects, classifies or tracks what appears in images and video.

Fits: When a task depends on looking: counting, checking, noticing.
Needs: Representative images, suitable cameras and lighting, and a defined acceptable error rate.

Software that carries a defined process from one step to the next without manual hand-offs.

Fits: When the steps are known and repeat, but still depend on someone remembering them.
Needs: A clearly mapped process and explicit rules for exceptions.

Systems that rank options for a person or a situation based on past behaviour and attributes.

Fits: When there are many options and useful history about what was chosen.
Needs: Enough interaction data, and a clear measure of a good recommendation.

Models that estimate a future value or outcome from historical data.

Fits: When past records exist and an estimate would change a decision.
Needs: Sufficient reliable history and honest evaluation on data the model has not seen.

An organized, searchable layer over what an organization knows, with access control.

Fits: When knowledge is scattered across documents, drives and people.
Needs: Agreed sources of truth, permissions, and a way to keep content current.

Pipelines and dashboards that turn operational data into a current, readable picture.

Fits: When data exists but reporting takes manual effort each time.
Needs: Access to the source data and agreement on the decisions it should support.

Start a conversation

Have a technology requirement, product idea, workflow or problem worth solving?