Agentic AI engineer: agents a company can trust

I build AI agents that do real work inside regulated companies: they read the evidence, cite every answer, check their own work and leave the decision to a person. I take on agentic projects as a freelance engineer or in a permanent role, in Paris or remote across EMEA.

oelhachi@gmail.com Download CV

What makes an agent usable in a company

An agent that answers fluently is easy to demo. In a bank or an insurer it also has to show where each answer comes from, follow a rule that someone can read, stop when it is unsure, and leave a trail that an auditor can replay.

That is most of the work, and it is the part I build.

Agents I have built

A-Pad: agents assess a project's documents against the criteria of its governance gate; every answer is cited and verified, an explicit rule gives the verdict, a person approves, and every run is logged. Six governance processes run on one platform.
KASTOR.AI: agents find the evidence in a company's own policies and draft the compliance-report sections the client chooses to delegate (GDPR, EU AI Act, NIS2, ISO 27001); a person signs.
Compliance assessment of AI coding tools: agents pre-fill the answers from public evidence in parallel; experts answer the rest.
A gate for agent extensions: every skill, prompt or tool server that coding agents install passes a security gate, and anything that runs code meets a reviewer.

Tools I use

LangGraph and LangChain, MCP for tools, retrieval over the company's own documents with hybrid search and re-ranking, automated evaluation with Ragas and DeepEval, Python and FastAPI, PostgreSQL with pgvector, models through Amazon Bedrock or Vertex AI.

Proof: the case studies

A-Pad

Agentic governance: cited verdicts, human override, audit trail.

KASTOR.AI

Compliance reports drafted by agents, approved by people.

Agent extensions catalogue

A security gate for what coding agents install.

Questions

What is an agentic AI engineer?

An engineer who designs and builds AI agents: systems where a model plans steps, calls tools and works on documents or data, with the controls that make the result reliable.

Which agent frameworks does Omar El Hachimi use?

Mainly LangGraph and LangChain, with MCP for tools, retrieval over the company's own documents (RAG), and automated evaluation.

Is he available for agentic AI projects?

Yes, as a freelance consultant or in a permanent role, in Paris or remote across EMEA. Contact: oelhachi@gmail.com.