AI Engineer and product builder based in Morocco
I build resource-efficient AI systems that move from papers and prototypes to useful products.
Portfolio · Research Notes · Apify · Substack · LinkedIn
My work sits at the intersection of autonomous agents, memory, retrieval, and production data infrastructure.
- Agent memory: episodic, semantic, and procedural memory that gives agents useful context without wasting tokens.
- Retrieval systems: methods for finding conceptual gaps and measuring whether retrieved context actually covers a question.
- AI-ready data: dependable extraction pipelines for RAG, model training, research, and market intelligence.
- Agent tooling: MCP integrations and automation systems that connect models to real business tools.
- Efficient AI: practical ways to run capable systems under compute, memory, and latency constraints.
| Project | What it does |
|---|---|
| CogDB | A unified memory database for AI agents, combining vector search, temporal knowledge, and reusable procedures behind one interface. |
| TopoRAG | Uses topological data analysis to detect conceptual gaps and blind spots in retrieved context. |
| n8n Marketplace Research | Turned 6,000+ public workflows into market analysis, an ML dataset, and a fine-tuned Llama workflow generator. |
| Apify Actors | 20+ production data products for automation intelligence, AI datasets, Moroccan markets, newsletters, and audience research. |
| Jira MCP Server | Lets AI assistants manage Jira issues, sprints, and projects through the Model Context Protocol. |
| Sentinel | An open negative-result study of a pump.fun trading strategy, with paper trading and on-chain evidence. |
I publish the reasoning behind the systems, including what failed and why:
- Research Notes - LLM infrastructure, agent coordination, memory, evaluation, and AI engineering.
- The Scraping Report - data-backed intelligence from automation and scraping ecosystems.
- Medium - the journey from 6,000 workflows to a fine-tuned and deployed model.
Python, TypeScript, PyTorch, FastAPI, Next.js, PostgreSQL, Docker, Apify, Crawlee, Playwright, n8n, MCP, and Linux.
I choose the stack around the system: model behavior, data contracts, observability, deployment constraints, and the people who need to operate it.
I am interested in AI engineering roles and collaborations involving agent infrastructure, RAG, model evaluation, data products, and production automation.
For custom Apify Actors, AI-ready datasets, MCP integrations, or an unusually difficult systems problem, reach me at mustaphaliaichi@gmail.com.