Amr K.

I am an AI/ML Engineer specializing in the end-to-end lifecycle of production-grade AI systems, with a core focus on Speech AI (ASR and TTS), Large Language Models (LLMs), and Agentic architectures. I thrive at the intersection of deep technical research and product engineering, building scalable AI solutions that deliver measurable business impact.

Currently working as an AI Engineer at Sawt-SA, I develop and deploy state-of-the-art Arabic (Saudi) speech models designed for robustness in noisy, real-world telephony and conversational environments. My technical expertise spans the entire AI pipeline:

Speech Technologies: I have trained natural-sounding TTS and highly accurate ASR systems, managing everything from data curation to optimized production deployment. In a previous role at Hams-AI, I optimized a speech recognition model to be over 30% more accurate, 90% smaller, and 12x faster.

LLMs & Agentic AI: I have managed major improvements in LLM function-calling capabilities and seamlessly integrated these models into production AI agent workflows. Additionally, I have built advanced Retrieval-Augmented Generation (RAG) systems, including Graph-based RAG and multilingual retrieval pipelines.

Evaluation & Deployment: I design custom in-house benchmarking pipelines and task-specific test sets to systematically guide model iterations. I actively deploy high-performance LLMs for low-latency inference using vLLM, FastAPI, Docker, and cloud platforms such as AWS and GCP.

I am currently advancing my academic foundation by pursuing a Master's degree in Artificial Intelligence at Ain Shams University. Whether I am making trade-offs between model quality, latency, and cost, or containerizing systems for real-world environments, I am driven by the challenge of making AI reliable, scalable, and highly performant.