Mahmoud E.

Key Skills

Software

No software listed

Top Subject Matter

E-commerce customer support and sales assistance using RAG/LLM systems

Industrial safety monitoring (PPE compliance) with object detection

Legal Services & Contract Review

Top Data Types

Text

Image

Document

Top Task Types

Question Answering

Object Detection

Freelancer Overview

Real-Time Industrial Safety Monitoring System — Computer Vision (2024). Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include LangChain, ChromaDB, and Groq. Education includes Diploma, Dr. Mostafa Saad (2025) and AI & Machine Learning Graduate, Digital Egypt Pioneers Initiative (DEPI) (2024). AI-training focus includes data types such as Text and Image and labeling workflows including Question Answering, Object Detection, and Fine Tuning.

Labeling Experience

AI Engineer, GenAI & RAG System - E-Commerce AI Platform

TextObject Detection

Built an intelligent sales agent and RAG-based e-commerce support system using modern LLM tooling and vector search. Owned the design of a dual-retrieval pipeline that combined semantic retrieval with live inventory data to improve autonomous resolution. Delivered production-ready, measurable low-latency performance with multi-provider LLM orchestration and containerized deployment. • Implemented RAG with LangChain and ChromaDB vector search • Integrated PostgreSQL inventory queries into retrieval for grounded answers • Optimized inference latency using Groq APIs and prompt engineering • Deployed a Dockerized backend to Hugging Face Spaces with provider fallback orchestration

2024 - 2025

Intelligent Sales Agent & RAG System — E-Commerce AI Platform (2024–2025)

TextQuestion Answering

Built and deployed an LLM-enabled e-commerce intelligent sales agent using a RAG pipeline that retrieves relevant knowledge and generates support/answer outputs for user inquiries. Fused vector search (ChromaDB) with live PostgreSQL inventory queries to support accurate retrieval-augmented responses during support resolution workflows. Implemented multi-provider LLM fallback orchestration to sustain response generation quality under provider rate-limit events. • Created dual-retrieval RAG architecture combining ChromaDB vector search with PostgreSQL inventory lookups. • Authored and optimized prompts for LangChain-based retrieval and response generation using Groq for low-latency inference. • Integrated Gemini/Claude APIs with fallback logic to preserve uptime and response reliability. • Delivered measurable performance improvements including autonomous resolution rate and reduced inference cost/token usage.

2024 - 2025

Computer Vision Engineer - Industrial Safety Monitoring System

ImageObject Detection

Engineered a real-time industrial safety monitoring system leveraging computer vision for PPE compliance detection. Developed and fine-tuned YOLOv11 models on a custom labeled dataset to detect hard-hats and safety vests in video streams. Optimized edge inference throughput and low per-frame latency using ONNX Runtime and efficient buffering. • Fine-tuned YOLOv11 with an annotated dataset for PPE compliance • Exported and accelerated models with ONNX Runtime for CPU edge deployment • Tuned video pipeline to sustain 28+ FPS with multi-threaded buffering • Implemented alerting and audit logging of bounding-box non-compliance metadata

2024 - 2024

Real-Time Industrial Safety Monitoring System — Computer Vision (2024)

ImageObject Detection

Engineered a computer vision PPE compliance pipeline by training YOLOv11 on a custom annotated image dataset of hard-hats and safety vests. Performed dataset annotation-driven fine-tuning to enable real-time detection and compliance decisioning within an edge deployment. Exported the trained model via ONNX Runtime and optimized inference to maintain high frame rate while producing timestamped compliance metadata. • Fine-tuned YOLOv11 using a 2,000-image hard-hat and safety-vest annotated dataset. • Implemented real-time video inference with ONNX Runtime export and multi-threaded buffering. • Produced detection outputs with low false-alarm rate and logged bounding-box non-compliance metadata. • Supported edge AI deployment constraints (CPU-constrained hardware) while maintaining throughput.

2024 - 2024

Education

Work History