Senior AI/ML Engineer with 4+ years of production experience building end-to-end AI systems — from RAG pipelines and multi-agent architectures to real-time voice agents and GPU-optimized inference services.

Currently based in Dammam, Saudi Arabia.

I build AI systems that actually ship to production — not just prototypes. My work spans the full lifecycle: architecture, training, optimization, deployment, and monitoring.

• LLMs & Generative AI — RAG pipelines, multi-agent systems (LangGraph, CrewAI), LLM fine-tuning (LoRA, QLoRA), prompt engineering
• MLOps & Infrastructure — CI/CD for ML (GitHub Actions), Docker, AWS (ECS, EC2, Lambda, SageMaker), model serving, auto-scaling
• Computer Vision — Object detection & tracking (YOLO, ByteTrack), segmentation (SAM), pose estimation, video analytics
• Voice AI — Real-time voice agents with Whisper (STT), custom TTS, VAPI, Twilio/Telnyx integration
• Backend — FastAPI async services, Redis caching, request batching, <200ms p95 latency at scale

No-Code Multi-Agent Builder Platform — Built a platform where users create and deploy custom AI agents without coding. Deployed on AWS ECS with Docker, auto-scaling, and load balancing handling concurrent users with sub-second latency.

High-Throughput LLM Inference Service — Designed FastAPI async endpoints with request batching and Redis caching achieving 10x latency reduction and <200ms p95 latency under load.

LegalMind — Modular RAG system for legal document Q&A featuring hybrid search (vector + BM25), Cohere cross-encoder reranking, mandatory citation validation, and AI evaluation agents (adversarial test generation, hallucination detection, citation verification) with CI/CD pipeline.

Real-Time Sports Video Analytics — Custom YOLO training for player/puck tracking in ice hockey. Optimized from 8 FPS to 25+ FPS via TensorRT quantization and GPU optimization.

Real-Time Voice Agent — Built calling agent using Twilio/Telnyx with Whisper for STT and custom TTS. Sub-1s end-to-end latency with VAD-based turn-taking and interruption handling.

Stable Diffusion Production APIs — Fine-tuned on custom datasets. Built production serving with request queuing and horizontal scaling for face restoration, AI aging, and image editing.

LLMs & GenAI:     LangChain, LangGraph, CrewAI, MCP, LlamaIndex, OpenAI, Claude, Gemini
                  Fine-tuning: LoRA, QLoRA, PEFT | Models: Llama, Mistral, Gemma
RAG:              Pinecone, ChromaDB, Qdrant, Weaviate, FAISS, Cohere Reranking
Voice AI:         VAPI, Twilio, Telnyx, Whisper, Deepgram, ElevenLabs, Coqui TTS
Computer Vision:  YOLO, ByteTrack, SAM, TensorRT, OpenCV, Stable Diffusion
MLOps:            Docker, GitHub Actions CI/CD, AWS ECS, Model Serving, Auto-scaling
                  vLLM, ONNX, TensorRT, Prefect
Cloud:            AWS (ECS, EC2, S3, Lambda, SageMaker, CloudWatch), GCP
Backend:          Python, FastAPI, Node.js, REST APIs, WebSockets, Async Programming
Data:             PostgreSQL, MongoDB, Redis, Vector DBs, Pandas, NumPy
ML/DL:            PyTorch, TensorFlow, scikit-learn, Hugging Face Transformers

• 🌐 Website: muaazdev.com
• 💬 LinkedIn: muaazdev
• 📧 Email: [email protected]