A Personal Project Journal | 2026 | Nidin Shah

Where I Started

I work as a technical specialist. I'm not new to technology — I understand systems, hardware, and infrastructure. I know my way around a terminal. I can read documentation without getting lost. But AI was always something I observed from the outside.

I used ChatGPT like most people do. I read about large language models, agents, RAG pipelines. I understood the terms but I had never actually built anything with them. There is a real difference between knowing what something is called and understanding it because you built it yourself.

That gap bothered me. So I decided to close it — by building a real AI project on my own machine, from scratch, over a few weeks.


What Made Me Start This

I wanted to build an AI agent — not just a chatbot that answers questions, but something that could take real actions. Search the internet on its own. Do math. Pull from knowledge bases. Remember what we talked about. All running locally on my hardware. No cloud. No API subscription. No usage limits.

I called it ARIA — Advanced Research Intelligence Assistant.


What is ARIA?

ARIA stands for Advanced Research Intelligence Assistant. It is an AI agent that I built and runs completely on my own pc (Localhost). No internet subscription needed to power it. No paying OpenAI or Anthropic per message. No data leaving my machine.

What ARIA do:

The most important thing about ARIA is that it decides which tool to use on its own. You just ask a question in plain language. ARIA figures out whether to search the web, check Wikipedia, or calculate something without you telling it what to do.

That decision making loop is what makes it an agent and not just a chatbot.


Installed Tools

Component Technology Purpose
LLM Qwen2.5 32B Core reasoning brain
Model Runtime Ollama Runs LLM locally on GPU
Agent Framework LangChain + LangGraph ReAct agent loop
Web Search DuckDuckGo API Real-time internet search
Knowledge Base Wikipedia API Deep topic knowledge
Calculator Python eval() Math operations
Memory Conversation history list Session context
Interface Python CLI User interaction
Chat UI Open WebUI via Docker Browser-based chat