Portrait of Asad Ijaz

Data scientist and automation developer, Faisalabad, Pakistan

I build predictive models and the software that puts them to work.

Asad Ijaz

About

I take a problem from raw data to something people can use: cleaning and exploring the data, training the model (churn prediction, recommendation systems), then wrapping it in an app or pipeline.

Lately that work is agentic. I build LangGraph systems that retrieve, grade their own answers and correct course, a voice assistant that works in Urdu and English with 35 tools, and pipelines that research, write and publish on their own.

I am a data scientist at Optimum Tech and a computer science student at the University of Agriculture, Faisalabad. I also take freelance work: ML models, RAG chatbots, data pipelines and automation.

Selected work

Telecom customer churn prediction

An end-to-end churn pipeline: SQL data normalization, exploratory analysis, model training in Python and an interactive Streamlit app that ranks at-risk customers.

  • Churn prediction
  • SQL
  • Streamlit

Source on GitHub

Agentic RAG with hybrid search and reranking

A LangGraph agent that routes each question, retrieves with dense plus BM25 search, reranks with a cross-encoder, grades what it found and corrects itself when the answer is weak.

  • LangGraph
  • RAG
  • Reranking

Source on GitHub

JARVIS voice assistant

An Urdu and English voice assistant behind a wake word and PIN, driven by a LangGraph agent with 35 tools and RAG over local files.

  • LangGraph
  • Voice
  • RAG

Source on GitHub

Aura Accessories shopping assistant

An agentic RAG chatbot for a luxury fashion brand, built as a LangGraph ReAct agent running on Groq.

  • LangGraph
  • Groq
  • RAG

Source on GitHub

AI email dispatcher

A Streamlit app where a Gemini-powered LangGraph workflow writes, themes, illustrates and sends personalised bulk email.

  • LangGraph
  • Gemini
  • Streamlit

Source on GitHub

Autonomous LinkedIn publisher

A multi-agent pipeline that finds a trending data science topic, researches it, writes a post, fact-checks it and publishes it to LinkedIn.

  • LangGraph
  • Multi-agent

Source on GitHub

University timetable management system

A Flask and MySQL web app for scheduling university classes, with Admin, Teacher and Student roles, a 7-table schema and automatic clash detection.

  • Flask
  • MySQL

Source on GitHub

Deep learning projects

Six TensorFlow/Keras projects in one repo: MNIST with an ANN and a CNN, CIFAR-10, Cats vs Dogs with transfer learning, a tuned California Housing ANN and an LSTM next-word predictor trained on Kafka's The Metamorphosis.

  • TensorFlow
  • Keras
  • CNN
  • LSTM

Source on GitHub

Classical machine learning projects

Four scikit-learn projects in one repo: a heart disease classifier, insurance cost regression with random forests, an end-to-end housing price pipeline and a movie recommender.

  • scikit-learn
  • Pandas
  • Recommender

Source on GitHub

More on GitHub

prize_bond_manager

A desktop app (Python + PyQt6 + SQLite) to keep track of your prize bonds, record draw results, and let a win-checking agent find winners for you. No win is recorded until you approve it.

Python, updated 2026-11-04

portfolio

Personal portfolio website (Flask)

Python, updated 2026-10-28

rag-projects

Small RAG and LangGraph agent experiments

Python, updated 2026-10-26

data-science-projects

Exploratory data analysis and BI projects in Python and Excel

Jupyter Notebook, updated 2026-10-26

advanced-rag-chatbot

A retrieval-augmented chatbot built with LangGraph, LangChain, and Gemini. It features a FastAPI backend and web UI wired around a self-correcting RAG pipeline. It answers queries using uploaded documents, targeted websites, or live web search fallbacks—evaluating and correcting its own answers before returning them.

Python, updated 2026-09-30

LMS-Result-Scraper-and-Analyzer

Python-based ETL pipeline that uses Selenium to scrape student results from the UAF LMS portal and Pandas to clean data, calculate GPAs, and track historical academic performance.

Jupyter Notebook, updated 2026-06-27

Container images

agentic-rag-langgraph

An LLM-driven agentic RAG system built on LangGraph with hybrid search and a Streamlit UI.

Updated 2026-09-28

ai-email-dispatcher

LangGraph AI agent using Gemini to write and send personalized bulk emails with custom themes.

Updated 2026-09-22

enterprise-ai-agent

LangGraph AI shopping assistant using agentic RAG, and FastAPI.

Updated 2026-09-16

ai-linkedin-publisher

LangGraph multi-agent pipeline using LLM to research, fact-check, and auto-publish LinkedIn posts.

Updated 2026-09-16

Experience

Chatbots

Assistants that answer from real product and document data, and decide for themselves when to look something up instead of guessing.

  • LangGraph
  • RAG
  • Streamlit

Built: Aura Accessories shopping assistant, Agentic RAG with hybrid search and reranking, JARVIS voice assistant

AI agents

Multi-step agents that plan, call tools, check their own output and finish the job, from research and writing to fact-checking and publishing.

  • LangGraph
  • Gemini
  • Groq
  • Multi-agent workflows

Built: Autonomous LinkedIn publisher, AI email dispatcher, Agentic RAG with hybrid search and reranking, JARVIS voice assistant

Deep learning

Neural networks built and trained end to end, from CNNs and transfer learning to sequence models for text generation.

  • TensorFlow
  • Keras
  • PyTorch
  • LSTM

Built: Deep learning projects

Machine learning

Predictive models from feature work to evaluation, including customer churn prediction and recommendation systems.

  • scikit-learn
  • SciPy
  • MLflow
  • NumPy

Built: Telecom customer churn prediction, Classical machine learning projects

Data scienceSince Jul 2026

The whole lifecycle: cleaning and exploring data, modelling it, and turning the result into insight people can act on. This is my current role at Optimum Tech.

  • Pandas
  • NumPy
  • Matplotlib
  • Plotly
  • SQL

Built: Telecom customer churn prediction, Classical machine learning projects

Artificial intelligence

LLM applications with better retrieval: hybrid dense and BM25 search, cross-encoder reranking, and agents that grade what they found and correct themselves.

  • LangGraph
  • Hybrid search
  • Reranking
  • Gemini

Built: Agentic RAG with hybrid search and reranking, JARVIS voice assistant

Python

My main language for data work, automation and small web apps, from Selenium scrapers to Flask and Streamlit interfaces.

  • Python
  • Selenium
  • Flask
  • Streamlit

Built: University timetable management system, AI email dispatcher, Telecom customer churn prediction

LaravelJan 2024 to Feb 2026

Professional web development on Laravel at Optimum Tech and IMPEKT Limited, before moving into data science.

  • Laravel
  • MySQL
  • SQL

Roles

Data Scientist

Optimum Tech, Punjab, Pakistan

Freelance AI & Backend Developer

Self-Employed, Remote

Laravel Developer

IMPEKT Limited, Pakistan

Laravel Developer

Optimum Tech, Punjab, Pakistan

Education

BS Computer Science

University of Agriculture, Faisalabad, Faisalabad

Skills

Data and machine learning

  • Python
  • Pandas
  • NumPy
  • SciPy
  • scikit-learn
  • TensorFlow
  • Keras
  • PyTorch
  • MLflow
  • Matplotlib
  • Plotly

LLM systems

  • LangGraph
  • LangChain
  • agentic RAG
  • hybrid search
  • reranking
  • ChromaDB
  • Streamlit
  • FastAPI
  • Gemini
  • Groq

Software

  • Flask
  • Laravel
  • C++
  • C#
  • .NET
  • SQL
  • MySQL
  • Selenium
  • Docker
  • Git

Let's work together

Open to data science, AI and automation projects. Send a message and Asad will reply by email.

Prefer email? mirzaasadijaz@gmail.com