Abdelrahman (Abdo) Mahmoud

I build computer vision and multimodal AI products, and the inference infrastructure that puts them into production.

Faces in moving cars. Tissue on pathology slides. Then the GPU serving layer and agent runtime underneath models like them. Thirteen years of teaching machines to read the physical world, and of finding out what it costs to keep that running in front of real customers.

The lesson repeats in every domain. Capability is the easy part. Unit economics, integration, regulation, and trust are the work.

Work

• DataRobot Principal Product Manager. The inference function: GPU inference, LLM and vision-language model serving, and the agent runtime for enterprise AI.
• PathAI Director of MLOps. Training and inference infrastructure for regulated digital pathology, supporting 60+ data scientists building deep learning models on tissue imagery.
• Affectiva (acquired by Smart Eye) Founding product manager for automotive AI. In-cabin perception on automotive-grade edge GPUs: occupant detection, gaze, drowsiness, cognitive load. Drove the pivot to safety-mandated occupant monitoring and a $2B category. Earlier, shipped the real-time facial expression SDK and the video data pipeline behind it.
• MIT Media Lab Affective Computing group. Mental state inference from facial video. The research that seeded Affectiva.

Every place we collapse a rich input into text to make it tractable, a quarter of the meaning is left on the floor.

Your Notetaker Heard Every Word and Missed the Meeting

Writing

On multimodal AI, perception systems, and the infrastructure that runs them in production.

• The next generation of AI-enabled cars will understand youIEEE Spectrum · co-authored
• A practical guide for platform teams managing shared AI deploymentsMay 2026
• Industry-standard LLM benchmarks in DataRobotMay 2026
• DataRobot + Nebius: an enterprise-ready AI Factory optimized for agentsMarch 2026
• Build enterprise-ready agentic AI with DataRobot using NVIDIA Nemotron 3 SuperMarch 2026
• How to build and scale your agent workforce with DataRobot and NVIDIAJuly 2025

All posts on the DataRobot blog

---

Also: 15+ patents in computer vision, machine learning, and automotive AI. Featured in IEEE Spectrum, TU-Automotive, and Autonocast.

If you’re building computer vision or multimodal AI products, or the infrastructure that runs them at scale, I’d like to hear from you. Find me on LinkedIn.