Applied AI · Multimodal Assistant
Multimodal AI Product Assistant
- Automotive
- Applied AI
- Selected prior engineering work by IYC's founder
- Anonymized
The workflow
A conversational product lookup needs to connect a customer’s question — typed, spoken, or a photo of a part — with catalog search and product details, in whatever language the customer uses.
The engineering challenge
Preserve conversation context across text, voice, and image input while exposing structured, multilingual product-lookup tools to the assistant.
The implementation
The chat service loads session history, invokes an agent, and stores the resulting messages. A dedicated tool searches the product catalog with structured, multilingual filtering (English, Arabic, and Arabizi) rather than free-text guessing, and a separate tool retrieves product details. Voice messages are transcribed before reaching the agent, and uploaded images are analyzed by a vision model so a photo of a part can drive the same lookup as a typed question.
Component overview
Technology
Technologies present in the project implementation.
Implemented capabilities
- Multilingual catalog search
- Structured search across vehicle, brand, category, and part-number filters, matching English, Arabic, and Arabizi terms for the same part.
- Voice input
- Spoken questions are transcribed before reaching the assistant.
- Image input
- Uploaded photos are analyzed by a vision model so a picture of a part can drive the same lookup as text.
Related work
Capabilities used: Multilingual catalog search, Voice input, Image input
Start
Have a system like this?
Tell us the hard part — we'll come back with a technical point of view.
Or email [email protected]