Skip to content

AI chatbot development services

AI chatbot development for your website, WhatsApp and Telegram

A useful AI chatbot answers from your own content, knows when to hand over to a person, and works where your customers already are: your website, WhatsApp or Telegram.

I build custom LLM chatbots on top of a proper retrieval pipeline. DocuChat is a multi-tenant chatbot platform with an 18 KB embeddable widget, WhatsApp and Telegram channels and a live human-handoff inbox. Thagi Alert's assistant verifies scam messages in 8 Indian languages.

Problems I solve

Generic answers

Off-the-shelf bots give vague replies. Your chatbot should answer from your documents, pages and policies, and cite them.

Customers on different channels

Some customers use your site, others WhatsApp or Telegram. One knowledge base can serve all of them.

Conversations the bot shouldn't own

Complaints, sales and edge cases need a human. Handoff to a live inbox keeps the customer in the same conversation.

Unknown cost and quality

Token costs and answer quality should be visible. I track cost per conversation and run quality evaluations.

What I build

Knowledge base
Ingestion from PDFs, documents and website crawls, kept up to date as your content changes.
Embeddable web widget
A lightweight script tag for any website, styled to your brand.
WhatsApp and Telegram
Bots on the Meta WhatsApp Cloud API and Telegram Bot API, sharing the same knowledge base.
Human handoff
A live inbox where your team takes over conversations, with the full history visible.
Multi-language
Answers in your customers' languages, including Hindi and other Indian languages.
Analytics and evaluation
Cost per conversation, unanswered questions and hallucination checks, so the bot gets better over time.

Related work

  • Screenshot of Jawably, ai voice receptionist for indian businesses, built by Anshuman Verma

    AI voice receptionist for Indian businesses

    Jawably

    An AI receptionist that answers business calls in Hindi, English and Hinglish, books appointments into Google Calendar and confirms on WhatsApp, with sub-second latency.

    Read the case study →

  • Screenshot of Navyora, headless shopify hydrogen store & ai commerce app, built by Anshuman Verma

    Headless Shopify Hydrogen store & AI commerce app

    Navyora

    A headless Shopify store for an Indian artisanal brand with a bilingual AI shopping assistant, plus an embedded Admin app for AI copywriting, review summaries and checkout extensions.

    Read the case study →

  • Screenshot of DocuChat, multi-tenant enterprise rag chatbot, built by Anshuman Verma

    Multi-tenant enterprise RAG chatbot

    DocuChat

    A multi-tenant RAG chatbot platform with WhatsApp, Telegram, live human handoff and an 18 KB embeddable widget.

    Read the case study →

How we'll work

  1. Step 1

    Collect the content

    Documents, FAQs and pages the bot should know, plus the questions customers actually ask.

  2. Step 2

    Build retrieval and answers

    A RAG pipeline with citations, tested on real questions before launch.

  3. Step 3

    Connect the channels

    Web widget first, then WhatsApp, Telegram or your helpdesk.

  4. Step 4

    Launch and improve

    Handoff, analytics and monthly review of unanswered questions.

Technology

  • OpenAI
  • Claude
  • Gemini
  • pgvector
  • LiveKit (voice)
  • Sarvam AI
  • Plivo SIP
  • Twilio
  • Laravel
  • Vue
  • Next.js
  • FastAPI
  • WhatsApp Cloud API
  • Telegram Bot API
  • HMAC webhooks

Frequently asked questions

Can the chatbot be trained on my website and PDFs?

Yes. The chatbot retrieves from your documents and crawled pages at answer time, so it stays current when you update content without retraining a model.

Can it work on WhatsApp?

Yes, through the Meta WhatsApp Cloud API, alongside your website widget and Telegram, all using the same knowledge base.

What happens when the bot can't answer?

It says so instead of guessing, and can hand the conversation to a person in a live inbox with the full history.

Does it support Hindi and other languages?

Yes. Modern models handle Hindi and other Indian languages well; Thagi Alert serves 8 Indian languages.

Who owns the chatbot and the data?

You do. It runs on your accounts and infrastructure, and you receive the code and documentation at handover.

Tell me what you're building

A short description of the problem and your data is enough. I'll reply with questions and a time for a call.

Start a project