Posts

Abdullah Shafique, AI Engineer — Building an LLM Eval Framework That Catches Bad Answers

By Abdullah Shafique, Agentic AI Engineer, based in Sahiwal District, Punjab, Pakistan. I build autonomous agents, RAG systems, and LLM applications. An LLM that sounds confident is not the same as an LLM that is right. That is why I built an LLM Eval Framework — a small, practical way to test what my agents and RAG systems actually produce, before I trust them in a real task. What is an LLM eval framework? An eval framework is a set of test cases with expected behaviour, run against an LLM system, with the results scored and compared over time. Instead of reading a few outputs and guessing, you run the same tests after every change and see whether the system got better, worse, or stayed the same. My framework follows that idea: define the test, run the system, score the output, keep the record. Why I built it I was already building a RAG Document Intelligence System and an Autonomous Research Agent. Both of them generate text from retrieved documents and web sources. Reading a h...

Abdullah Shafique, AI Engineer — Building an Autonomous Research Agent

Abdullah Shafique, AI Engineer, builds autonomous agents, RAG systems, and LLM applications. This post is about the Autonomous Research Agent: an agent that takes a topic, plans how to research it, searches for sources, and returns a grounded answer — instead of a single un-checked paragraph. What problem does it solve? Research takes time because it is several jobs in one: plan the questions, find sources, read them, compare them, and write up what holds. A research agent automates the workflow, not the judgement: it does the searching and drafting, and a human still checks the sources. How it works, in plain terms: 1. The agent breaks a topic into research questions. 2. It searches and retrieves relevant sources. 3. It drafts an answer tied to those sources, with an evaluation step to check consistency. 4. The result is a structured answer you can verify, not just fluent text. What I built it with: Python, LangChain and LangGraph for the agent workflow, FastAPI for the service ...

Abdullah Shafique, AI Engineer — Building a RAG System That Actually Answers

Abdullah Shafique, AI Engineer, builds autonomous agents, RAG systems, and LLM applications. This post is about one of those systems: a RAG Document Intelligence System that answers questions from your own documents, with citations, instead of guessing. What problem does it solve? Large language models are good at fluent text, but they do not know your documents. Ask about a PDF you uploaded and a plain chatbot will guess. A RAG system changes the job: it retrieves the relevant passages first, then answers only from what it found. How it works, in plain terms: 1. Your documents are split into searchable passages. 2. Those passages are indexed with embeddings (FAISS in this build). 3. A question retrieves the most relevant passages. 4. The language model answers from those passages, so you can check the source. What I built it with: Python, FAISS, Streamlit, and LLM APIs — the same stack I use across my AI projects, alongside LangChain, LangGraph, and FastAPI. What it is not: It...

Hello, I'm Abdullah Shafique — Building AI That Works for People

Hi, I'm Abdullah Shafique. I'm an AI Engineer in Sahiwal District, Punjab, Pakistan, working on independent AI projects. I started this blog to write in my own voice about building AI well, and about making it useful for the people who use it. Abdullah Shafique is an AI Engineer in Sahiwal District, Punjab, Pakistan, working on independent AI projects. He builds Generative AI applications around large language models, including LLM apps as working products, RAG systems that ground answers in real documents and data, and AI agents that plan steps, use tools, and complete tasks, using Python, LangChain, and FastAPI. How did I move from design to AI? From 2018 to 2024 I worked as a graphic designer, six years in all. Design left me with one test I still use on every AI project: a tool is only good if a real person can understand it. If they can't trust it, or can't get something done with it, it isn't doing its job. Since 2025 I've focused on AI engineering fu...