About
I'm an AI engineer and researcher based in Amsterdam, Netherlands. Currently at KPMG, I build AI-powered solutions ranging from healthcare AI assistants to global research platforms. My background spans computer vision, human-robot interaction, NLP, and full-stack development.
My research at Vrije Universiteit Amsterdam focused on sports analytics with computer vision and social robots in healthcare and education. I hold an M.Sc. in Artificial Intelligence from VU Amsterdam and a B.Sc. in AI from the Universiteit van Amsterdam.
Experience
AI Engineer & Full-Stack Developer
Working on two projects: a global health research platform (Node.js, Playwright, Azure DevOps) and a hospital AI patient-assistant avatar (Microsoft Foundry, React).
Research Assistant
The Dream Robot project: experiment design and AlphaMini behavior coding (Alpha SDK) for a social robot designed to deliver guided medical hypnosis to children during invasive procedures in pediatric hospitals.
Research Assistant
The Robot Bookworm project: a personalized AI reading robot using NAO. Contributed across the lifecycle (experiment design, dialogue pipeline, recruitment, data collection, qualitative + statistical analysis). The study showed significantly stronger book-related engagement and reader-book relatedness in the personalized condition.
Founder / Owner
Launched and manage an e-commerce platform specializing in traditional Pakistani clothing in the Netherlands.
WRApp Management
Managed WRApp, a workflow management application, improving efficiency across departments. Experience with Azure DevOps environment.
Education
M.Sc. Artificial Intelligence
Thesis: Real-time detection and tracking in sports analytics using computer vision. Research at the Sports Intelligence Lab and Theoretical Computer Science group. Published at ACM MMSports '25 and ACM/IEEE HRI '26. Minor in Entrepreneurship (2022–2023).
B.Sc. Artificial Intelligence
Thesis: bird's-eye-view transformation for a cyclist navigation app at Saivvy. Built a drone-captured BEV dataset and benchmarked Segformer, FCN, and PointRend on a 7-class urban segmentation task.