Kshitij Works

Independent AI products & engineering.
Gurugram, India.

← All products
Demo on request

Studio Desk

From a guest’s first question to their next order.

A restaurant AI assistant for websites and Telegram. Connect menu discovery, business questions, ordering, and staff support in one conversation.

A connected conversation
A guest starts a conversation WEB / TELEGRAM Studio Desk Business knowledge RETRIEVE Menu & ordering ACT Staff support HAND OVER
Workflow illustration

Designed for

Restaurants looking to connect customer conversations with their menu, ordering, and support operations.

The workflow

How it works.

  1. 01

    Start a conversation

    Guests ask questions through the website or Telegram, using text or voice in multiple languages.

  2. 02

    Find & order

    The assistant retrieves business information, helps guests explore the menu, manages a cart, and places orders through Petpooja integration.

  3. 03

    Keep staff in the loop

    Staff monitor conversations, take over a chat when needed, and explore sales through natural-language analytics.

Capabilities

Built for the whole workflow.

Business-specific answers

Retrieval-Augmented Generation (RAG) uses each business’s knowledge to inform responses.

Conversations with context

Persistent chat history and sessions support multi-turn conversations, with intent classification routing requests.

Connected ordering

Menu discovery, cart management, order placement, and order tracking connect the assistant to restaurant workflows.

Human support, included

Live chat monitoring, conversation history, and human-agent takeover share a management dashboard.

Getting started

Plan a demonstration.

A deployment needs business knowledge and menu information, channel setup, and ordering integration configuration. Contact me to discuss your restaurant’s workflow.

Enquire about Studio Desk ↗

For the technically curious

Behind the product

A multi-tenant Django platform combines LLM intent routing, FAISS retrieval, semantic caching, and persistent conversation context. Celery and Redis support asynchronous message processing.

Read the engineering case study ↗

Let’s talk

A different problem with similar constraints?

I also work with teams on custom AI products, integration, and production engineering.

Get in touch