Scraping
All public pages extracted to clean text: FAQ, returns, shipping, about, contact, every product page on arata.in.
Trained on all 95 SKUs, the full FAQ, return policy and shipping rules. Answers real customer questions in English, Hindi, and Hinglish.
95 SKUs trained · 5/5 adversarial tests passed · ~4 hours build time · 3 languages supported
Each was tested on the live agent before this page was published. Every answer cites real Arata policy or real SKUs.
All public pages extracted to clean text: FAQ, returns, shipping, about, contact, every product page on arata.in.
Brand voice and policy rules locked into the prompt. Science-first, no emojis, 3-day damage window, store credit only.
Personalised questions with real SKU names, run before the demo was shared. Includes 5 adversarial traps; all passed.
Hosting, embeddings, retrieval and rate limiting handled end-to-end. No custom infrastructure to maintain.
10 business days: kickoff call, full catalog retrain, Shopify theme embed, WhatsApp Business bridge, 50-question QA set, go-live runbook.
Get in touch