Mustafa Ayyub
Ai Engineer Intern @Deutsche Bundesbank
Signup · Get unlimited contacts
WORK HISTORY
Ai Engineer Intern @Deutsche Bundesbank
Frankfurt, DE
Developed a \"Visual RAG\" Pipeline: Engineered a multimodal retrieval system capable of answering questions from complex PDFs (tables, charts, and multi-column layouts) by treating pages as high-fidelity images rather than relying on error-prone OCR.Implemented Advanced Retrieval (ColPali): Integrated the ColPali model (ColBERT + PaliGemma) to generate visual embeddings, preserving the spatial context of documents often lost in standard text extraction.Optimized Inference Speed: Leveraged Groq’s LPU inference engine and Llama 3.2 Vision to achieve sub-second response times for visual question answering.Built End-to-End Architecture: Designed the full ingestion and inference workflow using Python and Byaldi, enabling the system to \"see\" and interpret documents similarly to a human reader.
EDUCATION
GC University Lahore
Bachelors in computer science, Computer and Information Sciences and Support Services
Bingen University of Applied Sciences
Master's in Computer Science, Information Technology
ABOUT MUSTAFA AYYUB
I am a Software Engineer and Master\'s student focused on bridging the gap between traditional software architecture and modern AI engineering.Currently, I am an AI Engineer Intern at Deutsche Bundesbank, where I am building multimodal retrieval systems that go beyond standard text parsing. Before this, I spent nearly two years at Devsinc and other firms building scalable full-stack applications, optimizing APIs, and managing complex integrations for fintech and CRM products.My core technical strength lies in shipping production-ready code whether that’s a RAG pipeline handling complex document layouts or a high-traffic Rails, Node.js backend. I am currently open to AI/ML and full-stack engineering roles where I can apply my experience in LLMs, prompt engineering, and system design.
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.