AI Developer

Overview

Designing and building a sophisticated Agentic AI application to provide in-depth, autonomous financial analysis for Small and Medium-sized Enterprises (SMEs). The system ingests and processes complex bank transactions and financial statements, using a team of AI Agents and RAG to identify trends, flag anomalies, and provide actionable insights in natural language format.

My Role

Lead developer for the entire system, from AI architecture to backend and frontend. This includes designing the agentic workflows and RAG pipelines.

Key Features

• AI Agent Team: Multiple specialized agents working collaboratively using LangChain framework
• RAG Pipeline: Retrieval-Augmented Generation for context-aware financial insights
• Multi-lingual Support: Utilizing the "Allam" open-source model for high-fidelity Arabic translations
• Transaction Processing: Automated ingestion and analysis of bank statements and financial documents
• Anomaly Detection: AI-powered trend identification and anomaly flagging
• Future-Ready Architecture: Designed for fine-tuning custom LLMs as the system evolves

Technologies

Overview

Led full-stack development of an AI chatbot specifically designed for industrial applications. The chatbot's primary purpose is to assist engineers by ingesting and understanding complex machine documentation, including PDFs and technical diagrams. Engineers use it to diagnose faults, understand machine alarms, and get immediate answers to technical questions.

My Role

Full-stack development - Built the Python/FastAPI backend and the responsive React frontend from scratch.

Key Features & Technical Innovations

• Advanced Vision RAG Pipeline: Processes both text and diagrams from technical documentation using Vision Language Models
• Anthropic's Contextual Retrieval: Implemented state-of-the-art contextual retrieval strategies for RAG, significantly improving answer accuracy and relevance compared to standard retrieval methods
• Multi-Modal Understanding: Extracts information from both textual content and visual diagrams in PDFs
• Real-Time Fault Diagnosis: Helps engineers diagnose machine faults and understand alarm codes instantly
• WebSocket Integration: Real-time streaming responses for immediate feedback
• Source Attribution: Cites specific sections of documentation for verification

Technologies

Overview

Instrumental in bridging AI initiatives with physical robotics operations. Successfully integrated a new Autonomous Mobile Robot (AMR) into the main production line and developed a vision-based dispatch system that autonomously routes the robot based on real-time visual input.

My Role

Responsible for designing and implementing the full integration between the AMR, factory management software, and AI vision system.

Key Features & Technical Implementation

• ROS2 Integration: Established robust, two-way communication between the robot's onboard ROS2 system and central factory management software via REST APIs
• AI Vision Dispatch System: Trained and deployed a YOLO model for real-time color detection at production stations
• Intelligent Routing: System identifies the color of a drone arm at a station, and based on this color, autonomously dispatches the AMR to its next correct location
• Production Line Integration: Seamless integration with existing factory workflows and systems
• Real-Time Decision Making: Vision model processes camera feed and makes routing decisions in real-time

Technologies