/ service / 05
AI development
LLM features, AI assistants, RAG search, automation flows, and model integrations inside product logic.
/ scope
What the work includes
/ stack
Technology focus
How 01.DEVS joins
We start with a short discovery, define risks, priorities, and the delivery plan, then move in iterations with transparent communication and a production focus.
/ questions
What to know before starting
Does a knowledge assistant need model fine-tuning?
Not necessarily. First evaluate retrieval, relevant context, document quality and expected answers. The choice between RAG, ordinary search and fine-tuning depends on the task and evaluation results.
How do you verify answer quality?
Create a set of typical and difficult questions with expected sources. Evaluate retrieval, evidence for answers, refusal when information is absent, permissions and execution cost separately.
Can we connect private documents?
Yes, after defining access rules, permitted providers and retention requirements. Authorisation must restrict documents before context reaches the model; hiding a source in the answer is insufficient.
Related insights
Demonstration
RAG knowledge assistant demo: a sourced answer or an honest refusal
An educational RAG example covering retrieval, citations, missing answers, access controls and evaluation without a paid public model API.
Read insight