YOUNIS B.

AI Engineer specializing in production LLM systems, Agentic AI, RAG, and Edge AI deployments. I build AI systems that go beyond prototypes — scalable, observable, and production-ready applications designed for reliability and real-world usage. My experience spans multi-agent systems, real-time Voice AI, advanced RAG pipelines, NVIDIA Jetson edge deployments, and computer vision systems, with a strong focus on latency optimization, workflow durability, and system observability. 🚨 Common AI Problems I Solve → Hallucinations and unreliable retrieval → Slow streaming and high-latency inference → Agents losing memory, context, or execution state → Poor observability and difficult debugging → AI demos that fail in production environments ✅ Production-Focused Solutions → Durable agent workflows using Temporal & DBOS → Advanced RAG pipelines with hybrid retrieval and reranking → Full tracing and observability for agents, tools, and LLM workflows → Real-time streaming architectures for Voice AI and conversational systems → Edge AI optimization on NVIDIA Jetson using TensorRT & DeepStream Core Expertise 🤖 Agentic AI & Multi-Agent Systems → Multi-agent orchestration and tool-use systems → LangGraph, LlamaIndex, AutoGen, Google ADK, PydanticAI → MCP servers, memory systems, planning, and HITL workflows 🔍 RAG & LLM Systems → Production RAG pipelines and knowledge-base AI → Hybrid search, reranking, embeddings, and retrieval optimization → GPT-4o, Gemini, Claude, LLaMA, Qwen, multimodal systems 🎙️ Real-Time Voice AI → Stateful voice agents with streaming audio pipelines → WebSocket-based low-latency architectures → Gemini-native voice agents and conversational workflows 👁️ Edge AI & Computer Vision → NVIDIA Jetson Orin/Nano deployments → YOLO, OCR, segmentation, tracking, and video analytics → TensorRT, DeepStream, edge inference optimization 🌐 Backend & Infrastructure → FastAPI, Django, REST APIs, WebSockets → PostgreSQL, Neo4j, Redis, vector databases → Docker, scalable deployment pipelines, event-driven systems ⭐ Selected Projects ✓ Conversena — Arabic conversational commerce AI with multi-agent workflows, hybrid RAG, multimodal search, and persistent memory ✓ VidEduGen — Multi-agent AI platform for automated educational video generation using scalable event-driven architecture ✓ Real-time Edge AI System — NVIDIA Jetson-based multi-camera vision pipeline with OCR, TensorRT, and DeepStream optimization ⭐ Why Clients Work With Me ✔ Production-first engineering mindset ✔ Strong expertise in Agentic AI and scalable LLM systems ✔ Real-world experience with Voice AI and Edge AI deployments ✔ Clear communication and maintainable system design If you’re building: → AI agents or multi-agent systems → RAG or knowledge-base AI applications → Voice AI or conversational systems → Edge AI / NVIDIA Jetson deployments → Computer vision or real-time AI pipelines I can help you design and ship a reliable production-grade AI system. 📩 Feel free to message me to discuss your project.