AI EVALUATION
Devometric.
Make retrieval quality something you can inspect.
RAGEvaluationsOllama
01
Prepare a dataset
02
Run retrieval
03
Evaluate answers
04
Review findings
01 / THE PROBLEM
Start with the friction.
A retrieval-augmented application needs a way to assess answer quality and examine how it behaves with untrusted input.
02 / THE APPROACH
Build a connected system.
An MIT-licensed RAG assessment platform with an evaluation harness, prompt-injection protection and support for airgapped runs using Ollama.
03 / THE OUTCOME
Keep the scope visible.
The project brings evaluation and local execution into the assessment workflow. This overview describes capabilities documented in the previous portfolio; no performance or accuracy benchmark is claimed here.
A SIMILAR CHALLENGE?
Talk to Risto