Price: 0
Number of applications: 10
06.08.26 (inclusive)
Pilot contract with CONTPARK LLC based on the selection results; volume and cost are agreed with the selected contractor
MVP
ICT tasks
Railway industry
Technologies in transport and logistics
Software/ IS
Multilingual AI consultants for complex B2B systems are difficult to test manually: the same question may give different answers in different languages, the model may add unconfirmed information or open an incorrect demo scenario. Without automated checks, updating the knowledge base and models increases the risk of factual errors, unsafe navigation, and quality degradation before demonstration to the customer.
Target effect: Automate at least 70% of recurring checks, identify critical discrepancies before release is published, reduce regression testing time, and receive an audited report on actual accuracy, route safety, language consistency, and response delay.
Vladimir Bondarev
Purpose and description of task (project)
To develop a module for automated quality and safety assessment of a multilingual voice AI consultant for industrial software. The module should check responses based on an approved knowledge base, the correctness of transitions according to permitted scenarios, compliance with the language of the dialogue, safe rejection in case of insufficient data, and response time. The result should be integrated with the current MVP of the AI consultant CONTPARK and generate a reproducible report on the quality of each release.
Note
The geography of the application is the whole world. The pilot is conducted on the CONTPARK demo circuit with test or anonymized data. The supported languages of the pilot are Russian, English, Spanish, French, German and Portuguese.
Whatsapp
