Available --:--local Start a project

( Blog · 18 articles )

Lab
notes.

Field notes on agentic engineering, AI automation and shipping production software with a team of agents.

  1. 01 AI automation for businesses in Saudi Arabia, the UAE, Egypt and Morocco: what to automate firstA practical guide to AI automation and AI agents for businesses in the MENA region: the first workflows worth automating, WhatsApp and Arabic-first design, data protection laws (PDPL, Law 09-08) and how to start small. 4 min read
  2. 02 Why every AI code reviewer needs a failing reproAI code review produces convincing noise. One rule fixes it: no finding counts without a failing test that reproduces it, plus a control that proves the test is valid. 2 min read
  3. 03 Arabic AI chatbot for business: understanding dialects and Darija, not just MSAHow to build an Arabic AI chatbot for business that handles Modern Standard Arabic, Gulf and Egyptian dialects, Moroccan Darija, Arabizi and voice notes, for customer support on WhatsApp and the web. 4 min read
  4. 04 Hiring a freelance AI developer in Morocco: a practical guide for companiesWhy companies in Europe, the Gulf and North America hire AI engineers and software developers in Morocco: nearshore time zones, languages, skills, costs, how to vet a developer and how to structure the work. 4 min read
  5. 05 Private LLM for businesses: local and self-hosted AI under PDPL, Law 09-08 and GDPRWhen a private, local or self-hosted LLM makes sense, what it costs you in quality and effort, and how data protection laws like Saudi Arabia's PDPL, the UAE PDPL, Egypt's Law 151 and Morocco's Law 09-08 shape the choice. 5 min read
  6. 06 How I ship production code with a team of AI agentsThe multi-agent pipeline I use on a production trading terminal: parallel implementer agents, hard quality gates, an adversarial reviewer, and a human sign-off. 3 min read
  7. 07 What to automate first: a practical guide to agentic automationNot every task should be handed to an AI agent. A simple way to pick the first workflows to automate, the patterns that work, and the mistakes to avoid. 2 min read
  8. 08 WhatsApp AI chatbot for businesses in Morocco: Darija, French and ArabicHow to build a WhatsApp AI chatbot with Darija AI for a Moroccan business: one that understands Darija, French, Arabic and voice notes: the official WhatsApp Business Platform, Meta's 2026 AI policy, handover to staff and data protection under Law 09-08. 4 min read
  9. 09 Claude Code subagents: a practical guide to writing your ownHow Claude Code subagents work, where the files live, which frontmatter fields matter, and how to design reviewer and implementer agents that actually help. 5 min read
  10. 10 RAG vs fine-tuning: which one your LLM project actually needsRAG vs fine-tuning explained by an engineer: when to retrieve your documents, when to fine-tune an LLM, when to combine them, and when better prompts are enough. 5 min read
  11. 11 Robotic bin picking with 3D computer vision: how it works end to endHow a vision-guided bin picking system works, from the 3D camera to the robot: segmentation, 6D pose estimation against CAD models, grasp selection and ROS 2. 5 min read
  12. 12 Multi-agent system architecture: practical patterns for LLM agents that workHow to design a multi-agent system with LLMs: orchestrator and specialists, pipelines, parallel workers, reviewers, shared state, and when one agent is better. 5 min read
  13. 13 How to build a RAG pipeline that works in production, not just in the demoA practical guide to building a RAG pipeline for production: document parsing, chunking, hybrid search, reranking, citations, evaluation, permissions and common failures. 5 min read
  14. 14 Demand forecasting for small businesses: a practical guide that starts simplePractical demand forecasting for small and mid-size businesses: baselines first, then statistical and ML models, honest backtesting, moving holidays and reorder decisions. 5 min read
  15. 15 LLM cost: practical ways to cut your API bill without losing qualityWhere LLM cost really comes from and how to reduce it: measure per task, right-size models, cache prompts and results, batch, trim context and run steady work locally. 5 min read
  16. 16 AI agent vs chatbot: what's the difference, and what is agentic AI?AI agent vs chatbot explained simply: a chatbot answers, an agent takes actions with tools. What agentic AI vs AI agent means, business examples, risks and costs. 5 min read
  17. 17 CatBoost vs XGBoost: which one to use for tabular data, and what matters moreCatBoost vs XGBoost in practice: categorical features, defaults, tuning effort, speed and overfitting, plus the validation mistake that matters more than the library. 5 min read
  18. 18 Faster Whisper: how to transcribe audio locally, on CPU or GPUA practical guide to faster-whisper for local Whisper transcription: model sizes, CPU vs GPU, int8 quantization, voice activity filtering and live transcription. 5 min read