Aleeza FATIMA

AI & Machine Learning Developer specializing in RAG and Conversational Agents

Based in: TOBA TEK SINGH, Pakistan

A Computer Science student with a strong foundation in Python, TensorFlow, and LangChain, focused on building intelligent systems like RAG chatbots and voice AI agents. She has successfully delivered projects ranging from disease classification models to real-time customer support solutions using modern DevOps and database tools.

Skills

Technical

machine learning

Python

Jupyter Notebook

VS Code

Streamlit

Pandas

TensorFlow

keras

Scikit-learn

deep learning

FastAPI

Dialogflow

Numpy

Git/Github

Seaborn

MySQL

Matplotlib

DSA

c++

Open cv

MangoDB

Gitlab

Docker

tool orchestration

ChromaDB

Pydantic

LangChain

Structured Decoding

Ollama

Quantization

GGUF

Languages

URDU

PANJABI

PUNJABI

ENGLISH

Projects

Aleeza's Kitchen - AI Ordering Chatbot

Developerpersonal

Built a conversational ordering chatbot for a restaurant using Dialogflow for intent recognition and entity extraction (menu items, quantities, orders). Developed a FastAPI backend to handle Dialogflow webhook requests and process user intents into structured actions. Designed a MySQL database and a db_helper.py module to store menu items, manage orders, and serve real-time data to the chatbot. Integrated the chatbot with a front-end website via webhook, using ngrok to tunnel the local FastAPI server during development.

PythonFastAPIMySQLDialogflowngrok

View the code

Potato Disease Classification

Developerpersonal

Built a CNN model to classify potato leaf diseases (Early Blight, Late Blight, Healthy) using the PlantVillage dataset. Trained and evaluated the model with TensorFlow/Keras, and developed a FastAPI backend to serve predictions.

TensorFlowKerasFastAPI

View the code

Estate Price Prediction Model

Developerpersonal

Built a logistic regression model using Scikit-learn. Applied train-test split, feature scaling, and evaluated model performance.

Scikit-learnPython

View the code

Voice AI Customer Support Agent

Developerpersonal

Built a real-time voice AI agent for telecom customer support that allows users to ask questions by speaking and receive spoken answers, rather than typing. The system converts speech to text, uses a LangChain-based agent to decide which action to take (searching plan details, checking network coverage, or looking up account balance), retrieves accurate information from a Qdrant vector database built from company documents, and converts the final answer back into natural speech. Key aspects: Implemented real-time speech-to-text and text-to-speech pipeline using LiveKit; Built an agentic RAG system with function-calling tools using LangChain; Parsed and indexed PDF documents using Docling and sentence-transformers embeddings; Configured natural conversation turn-handling to avoid interrupting users; Filtered internal AI reasoning steps so only final responses were spoken aloud.

PythonLiveKitLangChainQdrantGroqDoclingsentence-transformerstool orchestrationOllamaLlama 34-bit GGUF quantization

View the code

Telecom RAG Customer Support Chatbot

Developerpersonal

Built a Retrieval-Augmented Generation chatbot answering customer queries from three sources FAQ CSV, SQLite support tickets, and a PDF guide. Indexed data into ChromaDB using sentence-transformers embeddings, with recursive chunking for the PDF and per-record chunking for structured data. Built a merged multi-collection retriever and a LangChain generation chain with an LLM served via Groq, grounding responses to reduce hallucination. Delivered both a CLI and a Streamlit web interface for the chatbot.

PythonLangChainChromaDBGroqStreamlitsentence-transformersSQLiteCSV

View the code