Language: English
Request a service

AI assistants for support and commerce

Accuracy and cost audit of your AI chatbot or RAG app

Your AI assistant is live, but how often is it wrong? I build a labelled test set from your real questions, run it through your app and measure accuracy, made-up answers, refusals, speed and cost per answer. You get a clear report with the failures that matter and a prioritised fix list; in the higher packages I implement the fixes and measure again. Built by an engineer whose evaluation work cut wrong answers to unanswerable questions from 18 of 21 to 0 of 21 at 94.6% top-1 accuracy.

  • From $500
  • Delivery in 5–15 days
  • Fixed price, paid up front

See the packages

Accuracy and cost audit of your AI chatbot or RAG app

Choose a package

Starter

$500

Delivery in 5 days

  • 50-question labelled test
  • accuracy and made-up answer rates
  • fix list
Choose Starter

Standard

$1,200

Delivery in 10 days

  • 100-question test
  • cost and speed per answer
  • top fixes implemented
Choose Standard

Advanced

$2,500

Delivery in 15 days

  • 200-question test
  • retrieval and prompt redesign
  • before/after report
Choose Advanced

List prices in USD, excluding VAT. The amount in your payment currency, at the day's rate and with any VAT, is shown before you pay.

  • LLM
  • RAG
  • AI Evaluation
  • Chatbot
  • Quality Assurance

Order and pay

Your details are used for the invoice and to start the work. You pay on the secure page of the payment provider; card details never reach this site.

Package

The receipt and your order page link go here.

How it works

  1. Choose a package and pay online.
  2. I email you to ask for access and the details I need.
  3. The work starts once I have them; the delivery time counts from then.
  4. You get the result and a short written summary of what was done.

Need something that does not fit a package? Send a request