Abdulrahman Omar

Acting as Fractional CTO for Snippet, an AI-powered Zettelkasten knowledge-base platform designed for large-scale technical documentation and long-living knowledge systems. I lead the end-to-end technical strategy from translating business requirements into system architecture, to building and deploying production-grade AI pipelines. The core mission is to make knowledge bases correct, consistent, and evolvable as they grow. The core challenge Solving one of the hardest problems in knowledge systems: how to REPLACE, MERGE, or INSERT new information into an existing knowledge graph without breaking semantic consistency, citations, or hierarchical structure. This includes handling conflicting sources while preserving trust, traceability, and context. What I’m building? Architecting a hybrid AI system that combines deterministic guardrails (entity resolution, graph validation, cycle detection, citation verification) with modern LLM-based generation Designing conflict-resolution pipelines for multi-source knowledge ingestion Building graph-based semantic search with context-aware snippet retrieval Implementing evaluation frameworks for coverage, cohesion, redundancy, and drift Creating production-ready data pipelines for continuous knowledge base updates and maintenance Leading the full technical lifecycle: system design, implementation, deployment, monitoring, and iteration Validation & early adoption First client: EndeavourOS (20M+ users), validating the platform on real-world Linux technical documentation at scale. Tech Graph database neo4j · Knowledge graphs · RAG architectures · LLMs · Semantic embeddings · NLP pipelines. Our focus is similar in spirit to Perplexity, but applied to AI-powered knowledge bases rather than search engines. The competitive edge we’re developing is the fusion of deterministic guardrails with modern GenAI. Overview of the vision: https://www.youtube.com/watch?v=ILhyBQMdY50 Reference inspiration: https://www.perplexity.ai/hub

-Evaluated baseline RAG system with Ragas and tailored test datasets. -Architected and implemented an advanced RAG system for activeQ Protest’s RFP automation platform, improving accuracy by over 15%.

Developed a traveller-focused recommendation system at SIGMA EMEA by integrating customer segmentation (demographics, spending, device type, tenure) with geospatial tracking. Implemented tailored offers and insights to enhance personalization, improve customer engagement, and drive retention

Designed and deployed a Support Vector Machine (SVM) model to classify 25,000+ cat/dog images from Kaggle, achieving 99% accuracy (top 5% benchmark). Engineered a linear regression model to predict real estate prices using square footage and room metrics, reducing prediction error by 12% vs. baseline. Developed a K-means clustering algorithm to segment 10,000+ retail customers into 5 behavioral groups, enhancing targeted marketing strategies. Translated unstructured problem descriptions into technical designs, improving team workflow efficiency by 20%.

Modules: Deep Learning, Data Structures, Data Mining, Algorithm Design. AI, Software Engineering, Data Structure