01
Generative Engine Optimization SaaS
TopoRAG
A Chrome Extension SaaS that applies topological data analysis to LLM embeddings, ranking semantic content gaps competitors leave open.
TDA + pgvector + production billingAI Engineer + Product Builder
I build AI systems that move from research to real users.
I'm Mustapha Liaichi, an AI engineer at AYAutomate. I ship LLM products, agent infrastructure, automation systems, and the occasional unusual experiment that turns into a useful tool.
Based in Rabat, Morocco
Building for the open web
50+AI projects delivered
10M+records processed
61K+article views
1,000+community members
Selected systems
Products, protocols, and public experiments.
A cross-section of work spanning SaaS products, agent memory, marketplace intelligence, fine-tuning, and open protocols.
Visit TopoRAG
02
Cognitive database for AI agents
CogDB
An AI-native memory engine combining episodic, semantic, and procedural stores with token-aware retrieval, decay, and consolidation.
Adapters for AutoGen, LangGraph, and MCP03
Sales intelligence for automation freelancers
ParatusAI
Pre-meeting company briefings that surface automation opportunities, pain points, discovery questions, and ready-to-send outreach.
Native MCP server with persistent API keys04
Marketplace intelligence actor
Apify Store Analyzer
Analyzes public Apify Store data to uncover demand, pricing patterns, category momentum, competitors, and promising actor ideas.
Built from weekly analysis of 10,000+ actors05
Fine-tuned Llama 3 model
n8n Workflow Generator
A production model fine-tuned on thousands of real automation examples to generate structured n8n workflows from natural language.
8B parameters, Apache 2.006
Federated Agent Protocol
FAP Contribution
The first governed LLM-backed participant in an open federated coordination protocol, extended for Arabic, Darija, and French agents.
Open source, multilingual agent coordinationBuilder loop
Build it. Study it. Share what worked.
My public work connects product development with market research and education. I analyze automation ecosystems, turn the findings into tools, and publish the practical lessons for other builders.
Research notes
Learning in public, one system at a time.
Long-form explorations of LLM infrastructure, multi-agent systems, memory, AI engineering, and Morocco's emerging AI ecosystem.
Browse all 10 notesLLM Systems Deep Dive #7
35 Years of Agents - From BDI to AutoGen, and What We Lost Along the Way
Two papers published in 1991 disagreed on almost everything. The reactive versus deliberative debate set the terms for 35 years of multi-agent research.
LLM Systems Deep Dive #6
How I Built a Memory Database for AI Agents
A budget of 500 tokens with the right memories can beat 4,000 with random ones. This is the architecture behind CogDB.
LLM Systems Deep Dive #5
Morocco's AI Bet - Sovereignty, Darija, and the Infrastructure Nobody Has Built Yet
Morocco is building an AI ecosystem from scratch. But sovereignty requires systems that work for the languages people actually speak.
LLM Systems Deep Dive #4
AI Engineering Is Not Software Engineering - And We Don't Have the Tools Yet
Software 2.0 changes the job from writing explicit instructions to writing specifications of intent. Our engineering tools have not caught up.
Field notes
Writing from the workbench.
Market signals, automation patterns, experiments, and honest post-mortems.
Experience
From client delivery to enterprise AI.
AI Engineer
AYAutomate
Designing end-to-end AI automation for international clients, from RAG and multi-agent workflows to intelligent document processing and reusable agent architecture.
AI Engineering Intern
Lear Corporation
Deployed Palantir Foundry AI workflows for automotive requirements, reaching 89% accuracy and cutting manual review time by 75%.
AI Developer
Sweet Spot AI
Led AI development for a social intelligence platform processing millions of interactions, with RAG, GPT-4 reporting, OAuth, and billing.
AI Engineer & Consultant
Independent
Delivered 50+ AI projects for international clients and built data pipelines handling more than 10 million records.
Working set
Research depth.
Product instincts.
Production discipline.
Engineering degree in Data Science and Knowledge Engineering from the School of Information Sciences in Rabat.
01
AI systems
LLMs, RAG, multi-agent architecture, TDA on embeddings, prompt engineering
02
Agent tooling
MCP, LangChain, CrewAI, Mastra, AutoGen, LangGraph, Hugging Face
03
Product stack
Python, TypeScript, FastAPI, Next.js, Django, PostgreSQL, Supabase
04
Infrastructure
Palantir Foundry, Vercel, Railway, Docker, AWS, Azure, pgvector