Mohammed Shinoy

Building production AI infrastructure across Arabic NLP, speech, and multimodal systems. Led development of key subsystems for Fanar (Arabic LLM platform) and sole maintainer of the QARTA map routing engine.

• ▸ Built core components of Fanar — the region's first Arabic-native multimodal generative AI platform (RAG, ASR, Image Understanding, Mobile App).
• ▸ Engineered a high-throughput translation pipeline processing 30B+ tokens for LLM training.
• ▸ Designed an Islamic RAG system with <2s end-to-end retrieval time.
• ▸ Migrated vector DB from Chroma → Milvus for improved retrieval reliability; integrated with LangChain.
• ▸ Maintained QARTA routing engine at 99% uptime serving 70K+ API calls/day.
• ▸ Published the Fanar App on App Store and Play Store (React Native).
• ▸ Co-authored: "Fanar: An Arabic-Centric Multimodal Generative AI Platform".

Mentored students in the Machine Learning and Self-Driving Car Nanodegrees. Covered supervised, unsupervised, reinforcement learning, and deep learning — debugging student code and providing domain guidance.

Classified drowsy driving behaviour from facial images using TensorFlow. Built a data visualisation dashboard with Plotly/Dash from scraped research data. Implemented full-text search with ElasticSearch.

Built automated keyboard reliability tests on a Fanuc robot. Set up continuous integration for the team, accelerating test system deployments. Trained coworkers on Fanuc robot programming.