Casablanca · AI engineering & backend
I build AI that works on real data in production, and keeps that data safe.
AI engineer with a backend foundation: RAG and MCP over private data, Saudi-dialect speech, recommendations and privacy-aware assistants, evaluated before they ship.

What I build
See all projects →Agentic systems & AI harnesses
RAG pipelines with an MCP server as the retriever, and assistants that act with the user's own permissions, redact personal data and ask before they write.
Top projectCustomer Voice: RAG pipeline with an MCP retriever →
NLP / NLU
Public chatbots on Botpress, Rasa and an in-house engine, from the backend and chat widgets to their security, plus intent and entity extraction tuned per client across Arabic, French and Darija.
Top projectPublic chatbots on Botpress, Rasa and an in-house engine →
ML / DL & speech
Whisper large-v3 fine-tuned for Saudi-dialect calls, from audio preparation to evaluation, serving live transcripts and voice agents.
Top projectSaudi-dialect speech-to-text →
Clients I have worked for
What people I worked with say
“He naturally thinks like an architect, able to dive deep into complex issues, design solutions the right way, and build systems end-to-end.”
Featured case studies
All 13 projects →
Flagship
Osoul platform
Revenue as a Service for fine dining. A multi-tenant SaaS that captures the data restaurants usually lose and turns it into revenue KPIs.

AI
Saudi-dialect speech-to-text
Whisper large-v3 fine-tuned for Saudi dialect, with an audio pipeline for WhatsApp and Asterisk calls, behind live transcripts and tool-using voice agents.

AI
Customer Voice: RAG pipeline with an MCP retriever
A RAG pipeline over structured data, documents and customer interactions, with an MCP server as its main retriever and auth, privacy and retention built in.
Public Contributions
All public contributions →
AI StonesAn AI harness for Claude Code: Markdown skills, guardrail hooks and JSON settings that protect and improve how the model operates.
archioObject storage in Go with a multi-node, high-availability cluster and semantic search over stored files, built on Docling.
translatorLow-cost, context-aware translation: NLLB-200 for simple text, an LLM above a complexity threshold, behind a three-tier cache.
Tracking kitA Go/Gin user-recognition service and a TypeScript browser SDK, with every event HMAC-signed.
pg_inspectorAn MCP server that gives AI assistants safe, read-only access to PostgreSQL.
Core stack
All skills →Most read
All posts →- A Vision about Software Architecture StylesOct 2022 · on Medium
- How React works under the hoodAug 2021 · on Medium
- The Full Path into the API GatewaySept 2022 · on Medium





