I build AI agents, ML pipelines, and the software around them.
A Saudi M.S. Computer Science graduate seeking AI/ML engineering roles. I led AI development for a robotic rock-collection project at Aramco and now research LLM agents that operate real infrastructure at KAUST, from agents that fix network outages together to a framework for deploying multi-agent systems.
Research project · Prof. Shihada's lab, KAUST
Leading development of a network lab where LLM agents troubleshoot live outages on emulated and physical routers, firewalls, and servers, to measure how reliably AI agents can operate real infrastructure.
Designed experiments where agents with restricted access coordinate over A2A to restore service, and built a dashboard that replays every agent action to show where they fail.
Research project · Prof. Canini's lab, KAUST
Co-developed DMAS-Forge, a Go framework that deploys one multi-agent LLM application as a single container or as distributed services over HTTP, MCP, and A2A, with tracing to compare speed, cost, and memory.
Built finance, travel, and marketing multi-agent apps with RAG, memory, and tool calling to test it on realistic workloads.
Research project · Prof. Canini's lab, KAUST
Built a Flower and PyTorch federated-learning framework that runs in simulation and across Google Cloud machines, and showed that just 10% malicious clients collapse the standard method to about 10% accuracy.
Designed variational autoencoder aggregators that catch these poisoned updates as anomalies.
View codeOpen source · Longhorn Developers
An open-source Chrome extension that more than 60,000 UT Austin students use to plan their courses.
Partnered with UI/UX designers in Agile sprints to turn Figma designs into React features, and owned the sign-in flow and "What's New" page from design to release through code reviews and GitHub Actions CI/CD.
View codeLed AI development for a robotic rock-collection project. Trained PyTorch classifiers (ResNet, ViT, MobileNet) that identify rock types at over 97% accuracy, replaced manual logging with a YOLO and OCR pipeline, and delivered a Streamlit app for operators to track every sample.
Raised a fault-detection model's accuracy by 20 points by generating realistic synthetic 3D training data in Blender. Also benchmarked state-of-the-art PyTorch change-detection models on satellite imagery, which helped shape the scope of a new R&D project.
Mentored six student teams from first idea to a deployed full-stack web app on AWS (Python, Flask, React, MySQL), guiding them through planning, debugging, and deployment. Also built a Python autograder that cut grading to about five minutes per assignment, so students got feedback faster.
MS, Computer Science King Abdullah University of Science and Technology Expected
BS, Computer Science The University of Texas at Austin
Python, Go, TypeScript, PyTorch, React, Docker, PostgreSQL, and OpenTelemetry.
Email me about AI, machine learning, or software engineering roles and projects.
abdulrahman.alshahrani@kaust.edu.sa