Lead AI Engineer & LLM Specialist | PhD in Applied Mathematics
PhD-qualified AI engineer with 14+ years of experience spanning quantitative finance research and hands-on Generative AI engineering. I build agentic harnesses and architectures — tool orchestration, scoped capabilities, control loops — plus RAG pipelines and multi-agent systems for regulated industries (pharma, banking, transport), with deep expertise in evaluation methodology: building and meta-evaluating LLM-as-judge frameworks for audit-traceable environments.
- Agentic AI for Medical Writing — Autonomous document generation on a Skills Cards agentic harness: scoped MCP tool restrictions, DSPy pipelines, and streaming orchestration for pharmaceutical clinical trials
- Vision RAG — Multimodal retrieval systems that process tables, figures, and images from complex clinical documents
- Multi-Agent Orchestration — Exploring DSPy, CrewAI, and LangGraph for composable AI workflows
- Evaluation Harnesses & Anti-Hallucination — Building and meta-evaluating LLM-as-judge harnesses: claim-level scoring, calibration against expert annotations (Cohen's κ / Gwet's AC1), and cross-model verdict verification in production
| Project | Description | Stack |
|---|---|---|
| mcp-python-repl | Production-grade MCP server: persistent, sandboxed Python REPL with multi-session support | MCP, Python, Sandboxing |
| llm-auto-labeler | Multi-LLM consensus text labeling with mediation and explainability | Multi-LLM, Consensus |
| AWS MLOps | ML model development, deployment & monitoring pipelines on AWS | SageMaker, Step Functions, Lambda |
| LangChain + Vespa for Retail | RAG-powered recipe assistant & product recommendations | LangChain, Vespa, Hybrid Search |
| 🔨 Coming soon | Multi-agent orchestration benchmark: DSPy vs CrewAI vs LangGraph | DSPy, CrewAI, LangGraph |
I write about AI engineering, RAG systems, and agentic architectures on Medium.
- PhD in Applied Mathematics — Cadi Ayyad University, Morocco (stochastic differential equations, portfolio optimisation under stochastic constraints)
- Master MASEF — Université Paris-Dauphine
- Anthropic Claude Certified Architect — Professional (2026)
- Anthropic Claude Certified Developer — Foundations (2026)
- AWS Certified Machine Learning — Specialty
- Former quantitative strategist & analyst — Equity Risk Premia strategies and pricing models (Société Générale, Thomson-Reuters, RBC Dexia)
