ILYAS BELAOUD

– Architecting an Agentic AI pipeline to automate ADAS test case generation by processing multimodal requirements across various technical formats. – Streamlining safety validation by leveraging Multimodal LLMs to transform complex engineering documentation and visual data into executable testing scenarios. – Implementing autonomous agents to reduce manual overhead and enhance scenario coverage for safety-critical autonomous driving systems.

– Engineered a Hybrid Multi-Vector + GraphRAG system via LangGraph, merging semantic search with structured entity relationships; delivered an RBAC Admin Dashboard and the solution was selected for integration into ENSI's official website.

– Engineered a real-time detection + tracking pipeline (YOLOv8 + DeepSORT) with Kalman Filters, ROI counting, and confidence filtering for high-precision monitoring.

– Engineered a "Bid/No-Bid" decision engine using Fine-Tuned LLMs that cross-references tender criteria against company profiles, reducing assessment time by 60-90%. – Architected a robust data ingestion pipeline (Python + DeepSeek's OCR) to extract structured data (tables, numeric constraints) from large, non-standardized PDFs. – Deployed an Agentic RAG system that proactively interviews users to resolve missing compliance data, guiding them from document upload to final application readiness.

– Led the development of the AI Assistant for UMobility, a SaaS platform officially showcased at GITEX Africa 2025. – Orchestrated a Hybrid RAG system combining Elasticsearch and live APIs to deliver real-time, hallucination-free mobility data. – Engineered a tiered response pipeline that prioritizes cached answers, triggering LLMs only for novel queries to significantly reduce latency. – Implemented a dynamic scoring system using real-time sentiment analysis to automatically re-rank search results based on user satisfaction.

– Conducted a technical feasibility study of existing AI solutions for market analysis and automated business plan generation to define the roadmap for GenAI integration. – Customized and integrated MEAN Stack modules, refining business logic and adapting generic components to meet specific platform requirements. – Contributed to the system design phase by validating UML diagrams and suggesting data model optimizations prior to the implementation phase.