• 📋 Resume: Check out My Resume here
• 💬 Ask me about: Machine Learning, Deep Learning, Natural Language Processing & Computer Vision
• 📫 Reach me at: [email protected] | +20 1555171495
• ⚡ Fun fact: My real name is Mohammed El-Lithy but everyone calls me Panda 🐼

• MIT 18.06, Notes
• STAT 110, Notes
• MIT 6.006, Notes

Agolo · On-Site
Nov 2023 - Present

Key Responsibilities:

• Contributed to EntityLLM by fine-tuning TinyLlama on annotated English and German Wikipedia data for general/custom entity and relationship extraction.

• Developed a Warranty concern classification BERT model, achieving 86% F1 score; implemented with Gradio and deployed using CircleCI & ArgoCD.

• Fine-tuned an LLM to extract scenarios from unstructured warranty tickets (part, attributes, problems and conditions), utilizing MLFlow and returning structured data in JSON format.

Talents Arena Hybrid

Aug 2023 - Nov 2023 (4 months)
Location: Egypt

Key Contributions:

• Identified optimal embedding model for Semantic Similarity task among 19 English and 24 Arabic models to be used in production.

• Contributing in a Resume Parsing approach that efficiently extracts candidates information from CVs, organizing it into a structured format for database integration.

• Establishing a highly efficient Vector Database tailored to the company's requirements, carefully balancing cost-effectiveness, storage capacity, and security measures.

• Finetuning BERT with MLM and NSP (Domain Adaptation) on collected job posting data to achieve better context between job posts and their needed skills

Tanweer Remote

July 2023 - May 2024 (11 months)

Key Contributions:

• Finetuning of RWKV-4-World LLM for English to Arabic translation through data structuring and renting online GPU's.

• Conducted comparative analysis by fine-tuning various open-source models. Additionally, employed ChatGPT for facilitating multilingual translation tasks

• Transcription of multi language videos using WhisperX and deploying demo on GCP

DataSkew Hybrid

Mar 2023 - Nov 2023 (9 months)

Key Projects:

• Designed and executed a dynamic Time Series Forecasting project to predict the net profit for 8000+ products the company sells. Leveraged continuous data updates and iterative model training to enhance future predictions. Successfully deployed the forecasting solution on AWS

• Integration of an Arabic Chatbot into a university 's website of 13000 students and staff, using ChatGPT which incorporates the university's FAQ and private data. The Chatbot is then wrapped with FastAPI and deployed through Docker containers on GCP to be integrated with DialogFlowCX.

• Creating a women's health chatbot incorporating all the articles from the company's website as well as utilizing their private data. it is encapsulated using FastAPI and deployed using Docker on GCP. All user-bot conversations are stored in Firebase for easy retrieval and analysis of all 10,000 visitors monthly.

• A Sales Chatbot for an e-commerce and shopping website to engage with users on the specific web pages they are browsing, extracting relevant information about products to provide accurate answers to user inquiries and their follow-up questions in a Streamlit app

Description:
🚀 The goal of this project is to try a full CI/CD Pipeline with GitHub Actions and pre-commit hooks for a simple NLP Project to practice using Docker for Deployment and having Tests to make sure the new deployment is working as it should be

End To End Deployment

Business Goal:
🎯 Enable easy and simple prototype of a deployment process for any NLP model to be used in production

Tools & Technologies:

Description:
🐜 The aim of this project is to transform a large, general-purpose model into a Fast, Compact, and highly Accurate version Finetuned on a specific task, optimized for Low latency and suitable for Edge Computing. This involves Fine-tuning the model, applying model Distillation, Quantization, and leveraging ONNX Runtime (ORT) for efficient inference.

Finetune-Distill-Quantize-ORT

Tools & Technologies:

Description:
📜 This project demonstrates how to fine-tune a pre-trained model on a multilingual corpus and evaluate its performance across multiple languages, even on those not included in the fine-tuning process or languages with low resource availability.

Multi-Lingual-NER

Tools & Technologies:

Description:
💻 The goal of this project is to gain hands-on experience with PyTorch and explore its various capabilities, from basic tensor operations to advanced topics like distributed training, quantization, and model deployment.

PyTorch-All-In-1

Tools & Technologies:

Top Languages

GitHub Stats

Current Streak