AI applications

What is it?

An AI application puts a large language model (LLM) to work on one clear job: answering customers, searching company documents, or taking phone calls. Retrieval-augmented generation (RAG) lets the model answer from your own data instead of only what it learned in training.

Why use it?

A general chatbot does not know your products, policies or records. An agent built on your own data gives answers you can check, and it can be connected to your phone line, website and internal tools.

How I build it

Choose and configure the LLM for the job and the budget — OpenAI, the Claude API or Llama 3.
Index your documents and data with LlamaIndex so the agent can retrieve them (RAG).
Add phone calls or voice notes where needed — see Voice AI agents.
Tune the prompts against real conversations.
Serve it behind a FastAPI or Express API, deployed with Docker on AWS.

Why hire me?

I built Xobot (xobot.co) end to end, a no-code platform for custom AI agents with RAG and voice, and I work there as OpenAI & Prompt Engineer.

Related projects

Backend developer

AI-Powered Financial App

API for a mobile finance app with AI assistants trained on our dataset, voice-to-text through AWS Transcribe, OTP and social login, and PDF term sheets generated from Google Sheets and sent via SES.

Read the case study →

Express
TypeScript
OpenAI
AWS
MongoDB
Puppeteer

Blogs

Guides on AI applications

Claude API · 5 min read

How to build an AI agent with tool use on the Claude API (Python)

Build a Python AI agent on the Claude API: define tools with JSON Schema, run the tool-use loop, return results, and handle errors and stop reasons safely.

LLM · 6 min read

How to test an LLM app against real conversations

Test an LLM app by turning real, anonymized conversations into an eval set, scoring each reply with clear checks, and rerunning it on every prompt change.

LLM · 4 min read

OpenAI, Claude or Llama 3: how to choose an LLM for your product

How to choose between OpenAI, Claude and Llama 3 for a product: compare quality on your task, cost per completed task, latency, privacy and hosting, then test.

Prompt Engineering · 5 min read

Prompt engineering for production chatbots: a practical checklist

A production chatbot prompt needs a clear role and scope, grounding in retrieved context, a fixed output format, an "I don't know" rule, versioning and tests.

RAG · 5 min read

RAG vs fine-tuning: which one does your AI app need?

Use RAG when your AI app needs current facts and citations from your documents. Use fine-tuning when it needs a fixed style, format or narrow behavior.

RAG · 5 min read

What is RAG? A practical guide to retrieval-augmented generation

RAG lets an LLM answer from your own documents by retrieving relevant chunks at question time and adding them to the prompt. Here is how it works and fails.

FAQ

Questions about AI applications

Can you build a custom AI agent with the OpenAI or Claude API?

Yes. I build custom AI agents on OpenAI, the Claude API or Llama 3, and choose the model for the job and the budget. My guide “How to build an AI agent with tool use on the Claude API (Python)” shows the approach.

Can you build a RAG chatbot over our private documents?

Yes. I index your documents and data with LlamaIndex so the agent retrieves the relevant parts before it answers (RAG). Xobot, the no-code AI agent platform I built, connects different LLMs and data sources this way.

Can you add OpenAI or Claude to our existing SaaS app?

Yes. I add the AI features behind an API in FastAPI or Express, so your web app, mobile app and tools share the same data and rules.

How do you deploy an AI agent?

Behind a FastAPI or Express API, in Docker on AWS, with PM2 and Nginx. I test the prompts against real conversations before launch.

Have a bot, a backend or a strategy in mind?

Tell me what you want to build and where you are with it. Send a few lines about the project and I’ll reply with questions and next steps.

Rajshahi, Bangladesh

Message on WhatsApp Send an email