Portfolio | Yossra

Hello, I am Yossra!

About Me

Passionate about AI. My journey is centered on exploring cutting-edge NLP and LLM technologies, with a strong emphasis on real-world conversational systems.

Skills:

#LLMs #RAG #Agents #LLMOPS #GenerativeAI #LangChain #HuggingFace #Ollama #NLP #SpaCy #NLTK #MachineLearning #TensorFlow #ScikitLearn #CloudComputing #AWS #GCP #Azure #Docker #GitHubActions #WebDevelopment

Areas

LLMs

NLP

3. Agent4Debate-AR: Arabic Autonomous Debate Agent

Extended an existing self-debate framework inspired by the "Agent4Debate" paper to support debates in the Arabic language. Implemented an Arabic-specific debate flow, addressing linguistic and cultural nuances through custom prompt engineering. ]

Find in : https://github.com/Yossranour1996/Agent4Debate-AR

Conversational AI

4.Georgian Language Telegram Bot

Developed a Telegram bot to assist users in learning the Georgian language, with a specific focus on medical vocabulary to support healthcare professionals. The backend, built with GPT-4o.

Find in: https://github.com/Yossranour1996/Telegram-Assistant

5. VisionText RAG

This project is a Multimodal Retrieval-Augmented Generation (RAG) system that integrates text and image data for enhanced document understanding and content retrieval.

Find in : https://github.com/Yossranour1996/Multimodal-RAG

2. LLM-Twin

Created a personalized Large Language Model (LLM) Twin to reflect both professional and personal expertise by leveraging scraped data. The project utilized a three-pipeline approach : Data pipeline for scraping and preprocessing data, Model training pipeline for fine-tuning the LLM and Inference pipeline for interaction and continuous monitoring.

1. Google Dialogflow for Weather Bot Project

Developed an interactive weather bot using Google Dialogflow and Flask, integrating a weather API to provide real-time weather updates.

Find in : https://github.com/Yossranour1996/Weather-Bot-with-Google-Dialogflow

6. NLP Pipeline for Hate Speech Detection

Built a robust end-to-end NLP pipeline to detect and classify hate speech in textual data. The pipeline encompassed data ingestion, preprocessing, validation, model training, and evaluation stages.

Find in:https://github.com/Yossranour1996/End-to-End-NLP

7. RAG-with-Knowledge-Graph-DB

This project demonstrates the integration of Retrieval-Augmented Generation (RAG) with a Neo4j knowledge graph database and LangChain to enhance AI applications. This approach leverages Neo4j's graph database capabilities to store and manage interconnected data, enabling efficient retrieval of relevant information.

find in : https://github.com/Yossranour1996/RAG-with-Knowledge-Graph-DB/tree/main

8. Running-llama2-On-CPU

Provides a framework for deploying Meta's Llama 2 language model on CPU-based systems. This setup is particularly beneficial for users without access to high-performance GPUs, enabling them to utilize Llama 2's capabilities on standard hardware.

Find in: https://github.com/Yossranour1996/Running-llama2-On-CPU

9. SQL Agent

Developed a Text-to-SQL system using LangChain agents for dynamic SQL query generation and explanation, leveraging LLMs' function-calling mechanisms. Experimented with open-source models like Llama 3.1 (Ollama) and Qroq to evaluate and optimize the system's performance.

Find in : https://github.com/Yossranour1996/SQLAgent