Building AI where the internet doesn't reach.

Designed and shipped end-to-end Retrieval-Augmented Generation (RAG) pipelines over a 500GB+ multilingual technical documentation corpus — Arabic/English ingestion, ChromaDB vector storage, and Sentence Transformer embeddings.

Built and deployed Model Context Protocol (MCP) tools that let an internal LLM platform reason over crew readiness, aircraft configuration, and safety documentation — refactoring multi-query joins into optimized database views and significantly reducing latency.

Led a proof-of-concept custom chat UI to unlock MCP server elicitation (interactive user feedback during tool execution) where off-the-shelf OpenWebUI fell short.

Tuned LLM parameters (temperature, context window) and iterated on system prompts with stakeholders — measurably improving response consistency and tool-chaining accuracy.

Researched and piloted next-generation RAG techniques including RAPTOR (recursive tree-organized retrieval) and Knowledge-Graph RAG, proposing an incremental roadmap that preserves existing investment.

Debugged and resolved a token-starvation issue that degraded Arabic document retrieval — tracing the root cause across embeddings, context window, and tool-call loops.

Mentored junior engineers on MCP tool design, SQL query optimization, and RAG fundamentals.

Deployed AI systems inside an air-gapped, security-restricted intranet — coordinating with infrastructure, database, and integration teams across UAT and production environments.

Earned the Neo4j Professional Certification and began applying graph modeling to predictive logistics scenarios such as part-cycle tracking and aircraft serviceability.

Utilized the React framework to methodically plan, design, and construct design system components for an enterprise resource planning (ERP) application — a suite of 50+ precisely built components including DataGrid, DataLookup, DropDown, Slide Panel, and the application Shell.

Crafted UI components for dynamic dashboards (e.g. Operation Dashboard) — integrating interactive charts and tabular data, enhanced with fluid animations and a polished interface.

Designed, developed, and deployed a mobile application with seamless offline functionality for aircraft pilots and technicians to record mission info, flight hours, faults, and scheduled inspections — syncing with the central ERP using advanced encryption.

Built a Geographical Information System using open-source Apache Flex (Adobe Flex) to mark the location of aircraft as they are deployed — known internally as Fleet Location.

Created a server health check monitoring system using socket technology for continuous real-time data updates — disk consumption, CPU utilization, and memory usage with 24/7 availability.

Developed robust "File View" backend systems converting Excel, Word, and PowerPoint to PDF then SWF — enabling viewing without direct file downloads and protecting system security.

Built a client-centric reporting mechanism with React, encapsulated as a reusable component — expediting report development and freeing it from legacy Reporting Engine Server dependencies.

Developed a complex Scheduler component to render aircraft mission schedules (Weekly Mission) — drawing schedule blocks accurately in a columnar structure with high-performance rendering.

Actively assisted colleagues in implementing attractive UIs and recommended the right approaches for optimal front-end performance.