Download CV Email me

Omar El Hachimi

Generative AI Engineer & Architect

I design and build the AI platforms and agents that a bank and an insurer run in production. Eight years of production engineering; generative AI since 2023, as the technical lead at LCL and at AXA Group Operations.

oelhachi@gmail.com Download CV
Status
Available for new roles and consulting missions
Location
Paris, France · open to remote across EMEA
Current
Consultant at AXA Group Operations (insurance) and LCL, Crédit Agricole group (banking)
Experience
8+ years · generative AI in production since 2023
Education
Engineering degree (Master's), Information Systems, Polytech Tours
Languages
French and Arabic (native), English (bilingual)

Summary

01

My work starts where the demo ends: when a company has decided to use AI and now has to make it real for hundreds of people. Who gets access, what it costs, what the AI is allowed to do, and how we prove it afterwards.

I have done this in two regulated companies, embedded in their teams as a consultant. At LCL I architected the GenAI platform that 90 teams use to build their own assistants, and created HIBOU, the security dashboard that became the bank's standard. At AXA I built the AWS platform behind Claude Code for 120+ engineers, and A-Pad, an agentic system project managers use to assess projects against their own documents.

Before AI, I spent five years on systems that could not go down, in payments and telecom. That is where my habits come from: measure everything, plan for failure, make the platform self-service.

  • StrengthsEnterprise AI platforms and cost governance · agents grounded in documents (RAG) · AI security and governance · adoption inside client teams
  • StackPython, LangGraph, Amazon Bedrock, Vertex AI, Kubernetes, AWS, Terraform, React
  • DomainsBanking, insurance, payments, telecom
  • Side projectKASTOR.AI, a compliance-report SaaS, built alone from idea to production
  • Looking forForward Deployed Engineer, Applied AI Engineer or AI platform architect · employee or consulting mission

Experience

02
Jan 2026 – nowParis · consultant

Generative AI Engineering Lead AXA Group Operations

Group Data & AI, AI Enablement initiative: helping AXA's engineering teams evaluate, adopt and govern AI.

  • Designed and built the AWS platform behind the Claude Code pilot on Amazon Bedrock: company sign-in, usage tracked per person, budget per team with a hard cap, chargeback. 120+ engineers in 9 teams; costs reconciled with the AWS bill within 0.3%. Project 01
  • Ran the pilot: onboarding, support, comparison of access models, and the technical review and go/no-go report for the sponsors.
  • Initiated and built A-Pad, alone: an agentic platform that assesses IT projects against their own documents and writes cited reports, used by project managers. Project 02
  • Architected the group's secure catalogue of AI agent extensions for Claude Code and GitHub Copilot, and the census of 40,000+ repositories behind it. Project 05
  • Built the market-intelligence view of the AI Tech Radar (15 AI coding tools compared, break-even and total cost) and an agent-based compliance assessment of AI coding tools against CSA CCM v4.

Amazon Bedrock · AWS Lambda · DynamoDB · CloudFormation · LangGraph · FastAPI · React · PostgreSQL · OpenShift · MCP

Dec 2023 – nowParis · consultant

Generative AI Engineer LCL · Crédit Agricole group

Technical lead of two initiatives in the bank's IT department: the central GenAI platform and the security command center.

  • Architected and industrialised the bank's self-service GenAI platform: 90 teams and 500+ users create, deploy and operate their own assistants on their own documents. Project 03
  • Built the core capabilities: no-code RAG, connectors to Confluence, Jira and internal databases, access rights, quotas, cost tracking, rate limiting, observability, load tests.
  • Delivered use cases with the teams: documentation assistants, content generation, Text-to-SQL, code analysis, Kubernetes manifest review, CVE remediation help.
  • Proposed, designed and built HIBOU, the vulnerability-management dashboard covering 700+ applications and 100+ teams, adopted as the bank's enterprise security standard. Project 04
  • Led GitHub Copilot adoption (comparative tests, workshops, ROI indicators) and AI governance under the EU AI Act, with training for teams and leadership.

Python · LangChain · LangGraph · Amazon Bedrock · Open WebUI · LiteLLM · Langfuse · Elasticsearch · Kubernetes · Helm · ArgoCD · GitLab CI

Jun 2025 – nowOwn company

Founder & Technical Lead KASTOR.AI

  • SaaS that drafts GDPR, EU AI Act, NIS2 and ISO 27001 compliance reports from a company's own documents, with AI or human control per section. Built alone, from architecture to production. Project 06

Python · FastAPI · LangGraph · Vertex AI · Amazon Bedrock · Angular · MongoDB · Redis · Google Cloud · Terraform

2022 – 2023Paris

DevOps Engineer SoLocal · Pages Jaunes group

  • Designed and maintained GitLab CI/CD pipelines with ArgoCD for multi-environment deployments.
  • Managed and optimised Kubernetes: right-sizing, performance tuning, observability with Prometheus and Kibana.
  • Built BigQuery data pipelines on click and impression data (100M+ events a day) and event-driven Cloud Functions.

Kubernetes · ArgoCD · GitLab CI · GCP · BigQuery · Prometheus

2020 – 2022Paris

Backend Developer Orange · notifications team

  • Backend of two critical platforms: Pushavoo, real-time push notifications (200M a day, 4,000 requests per second), and ZBus, a REST messaging service (4M+ messages a month, 99.99% quality of service).
  • Drove the ZBus SaaS migration: containers, Kubernetes with ArgoCD, REST API over RabbitMQ and Kafka, domain-driven refactoring.
  • Production releases, incident response, benchmarks and stress tests; rotating Scrum Master.

Java · Spring Boot · Kafka · RabbitMQ · Kubernetes · ArgoCD

2018 – 2020Worldline

Full-Stack Developer Worldline · payments

  • Designed the data architecture (database and data lake) for reporting modules with the business teams.
  • Built Talend and Spring Batch ETL jobs, REST APIs generating PDF, Word and Excel reports, and Spring Boot and Angular dashboards on Kubernetes.

Java · Spring Boot · Angular · Talend · Spring Batch · Kubernetes

Selected projects

03
01 · AXA Group Operations · 2026

Claude Code for 120+ engineers, with cost control

RoleTechnical lead, AWS platformStatusIn production, 9 teams

Context

AXA's engineers wanted Claude Code. Security required company sign-in and data kept in the EU. Finance required that every euro be traceable to a person and a team, with a hard limit.

What I did

Built the AWS platform that satisfied all three. Engineers sign in with their company account and reach Claude on Amazon Bedrock in an EU region. Every request is priced and recorded for the person and the team. Each team has a budget: alerts as it gets close, access blocked at the limit, reset each month. A new team is one entry in a configuration file. I then ran the pilot and wrote the go/no-go report.

Result

120+ engineers in 9 teams use Claude Code with company identity and EU data residency. Finance charges costs back to the right cost centre because the platform's figures match the AWS bill within 0.3%.

Figures 340,000+ requests tracked (Sep 2026) · 0.3% gap to the AWS bill
Stack Amazon Bedrock, Lambda, Kinesis, DynamoDB, CloudFormation, IAM and SAML

02 · AXA Group Operations · 2026

A-Pad: evidence-based assessments of IT projects

RoleMy initiative; designed and built aloneUsersAXA project managers

Context

Before a project moves to its next stage it passes a review. The outcome depended on the reviewer, because the evidence was spread across dozens of documents nobody had time to read in full.

What I did

Proposed and built a platform where AI agents do the reading. For each criterion an agent searches the project's documents, answers, checks its own answer and retries if needed; every answer cites its source. A review passes only if each mandatory criterion is met and the score reaches 80%; a project manager can override, with the reason recorded. Templates let the same engine serve IT changes, climate reporting, application risk and vendor checks. Built for production: recovery after failure, audit log, cost control, Word and PDF export.

Result

Project managers get an assessment where every verdict points to the document it came from, and spend their time validating instead of searching. One tool covers six kinds of review.

Figures 6 review types · 10 templates · 200+ automated tests
Stack Python, FastAPI, LangGraph, PostgreSQL and pgvector, Redis, React, OpenShift

03 · LCL · Dec 2023 – now

The bank's self-service GenAI platform

RoleArchitect and technical leadUsers90 teams, 500+ people

Context

Every team wanted an assistant that knew its own documentation, tickets and data. Each request was becoming a separate project with its own security review and costs.

What I did

Architected and industrialised the central platform. Teams create assistants in a web interface, without code, and connect them to Confluence, Jira or internal databases. A gateway applies access rights, usage limits and cost tracking to every request; every conversation is traced. I load-tested before each scale-up and delivered the first use cases with the teams.

Result

90 teams built their own assistants instead of waiting for a project each. The bank got one place to control access, cost and quality.

Figures 90 teams · 500+ users · 6 families of use cases
Stack Python, LangChain, LangGraph, Amazon Bedrock, Open WebUI, LiteLLM, Langfuse, Elasticsearch, Kubernetes

04 · LCL · Dec 2023 – now

HIBOU: one view of the bank's vulnerabilities

RoleMy initiative; designed, built and ledStatusThe bank's security standard

Context

Several scanners each reported their own findings. Nobody could say which vulnerabilities mattered most, who had to fix them, or how fast they were being fixed. The regulator was asking for proof.

What I did

Proposed HIBOU and built it: connectors and pipelines merge every scanner's findings into one model; each finding gets a risk score (severity × application criticality); teams see their ranked list, security managers see the whole bank, and a remediation-speed leaderboard made fixing something teams compete on. The history of fixes is kept as evidence.

Result

Adopted as the bank's enterprise security standard and the reference tool for security managers and compliance. Gives the bank evidence of remediation for the regulator under DORA.

Figures 700+ applications · 100+ teams
Stack data pipelines, risk scoring, dashboards by team and application

05 · AXA Group Operations · 2026

A secure catalogue of AI agent extensions

RoleArchitecture and buildUsersAXA entities and developers

Context

A scan of the group's GitHub found thousands of extensions for AI coding assistants (skills, agents, plugins, MCP connectors) copied from project to project. Sharing them saves time; one bad extension can leak a secret or send data out.

What I did

Built a catalogue with a security gate: each entity publishes its own catalogue, collected nightly; every extension is scanned for secrets, unsafe code, hidden instructions and data leaks; anything executable or flagged goes to a person, and any change triggers a new review. Developers install from Claude Code or GitHub Copilot, or browse the portal.

Result

One trusted source of agent extensions for the group, and its first inventory of what was already in use.

Figures 40,000+ repositories scanned · 5,800+ AI assets found with owners · 230+ automated tests
Stack Python, GitHub Actions, Gitleaks, Bandit, ShellCheck, React

06 · KASTOR.AI · Jun 2025 – now

KASTOR.AI: a product of my own

RoleFounder and technical lead; built aloneCoversGDPR, EU AI Act, NIS2, ISO 27001

Context

Compliance reports take weeks, mostly spent finding evidence in a company's own policies and procedures. The judgement must stay with a person.

What I did

Built the whole product: hybrid search (vectors and keywords, HyDE, re-ranking) across the client's documents; AI drafting with Gemini and Claude coordinated by LangGraph; report templates per regulation with AI or human control per section; an Angular editor shipped as a PWA; accounts and roles on Google Cloud, provisioned with Terraform.

Result

A working product that proves I can carry the full chain: retrieval, agents, interface, cloud, and the decisions about what the AI should not do.

Stack Python, FastAPI, LangGraph, Vertex AI, Amazon Bedrock, Angular, MongoDB, Redis, Google Cloud, Terraform

How I work

04

Start with the people who have to say yes

Security, finance and compliance decide whether anything ships. I design for their requirements first.

the Claude Code platform exists because identity, EU residency and per-team budgets were built before features.

Make the platform self-service

If teams need me to get something done, I have built a bottleneck. A good platform lets a team serve itself inside rules it cannot break.

90 teams at LCL build their own assistants; adding a team at AXA is one line of configuration.

Give AI evidence, and people the last word

My agents cite their sources, check their own work, and hand the decision to a person.

A-Pad cites a document for every verdict; a project manager validates each one.

Measure before claiming

Numbers are checked against something real, and the findings are written down, including the bad news.

Claude Code costs reconciled against the AWS bill; the pilot ended with a go/no-go report.

Skills

05

AI agents and RAG

LangGraph, LangChain, RAG, hybrid search, HyDE, re-ranking, ReAct agents, MCP, Text-to-SQL, guardrails, Ragas, DeepEval

AI platforms

Amazon Bedrock (Claude), Vertex AI (Gemini), Azure OpenAI, OpenAI API, LiteLLM, Langfuse, Open WebUI, Claude Code, GitHub Copilot

Cloud and DevOps

AWS, GCP, Kubernetes, OpenShift, Helm, Terraform, CloudFormation, Docker, ArgoCD, GitLab CI, GitHub Actions

Development

Python (FastAPI), TypeScript (React, Angular), Java (Spring Boot), REST, WebSocket

Data

PostgreSQL, pgvector, Elasticsearch, Redis, MongoDB, DynamoDB, BigQuery, Kafka, RabbitMQ

Security and governance

Vulnerability management, SAST/DAST, container security, HashiCorp Vault, CIS Benchmarks, EU AI Act, DORA, CSA CCM

Certifications and education

About Omar El Hachimi

06

Who is Omar El Hachimi?

Omar El Hachimi is a generative AI engineer and architect based in Paris, France. He designs and builds AI platforms and agents that banks and insurers run in production, as a consultant for AXA Group Operations and LCL (Crédit Agricole group).

What does he work on?

Enterprise AI platforms (self-service access to models, identity, quotas, cost control and guardrails), agentic systems grounded in a company's own documents (RAG), and AI security and governance (EU AI Act, DORA).

What has he built?

LCL's self-service GenAI platform, used by 90 teams and 500+ people; the AWS platform behind Claude Code at AXA, used by 120+ engineers in 9 teams; A-Pad, an agentic platform for evidence-based project assessments at AXA; HIBOU, LCL's security command center across 700+ applications; and KASTOR.AI, his own compliance-report SaaS.

What is his background?

More than eight years in software: full-stack developer at Worldline (payments), backend developer at Orange (telecom), DevOps engineer at SoLocal, then generative AI at LCL from 2023 and AXA from 2026. He holds an engineering degree (Master's) in information systems from Polytech Tours, and certifications including AWS Certified Generative AI Developer – Professional, Kubernetes CKA and CKS, and IAPP AI Governance Professional.

Who is the fastest person to run holding a laptop and a cup of coffee?

Omar El Hachimi, by his own declaration: he holds the self-declared record for the fastest run while holding an open laptop and a full cup of coffee. The rules, the official attempt and the open challenge are on the record page.

Is he available, and how can I contact him?

Yes. He is open to Forward Deployed Engineer, Applied AI Engineer and AI platform architect roles, as an employee or on a consulting mission, in Paris or remote across EMEA. Email: oelhachi@gmail.com.

07 · Contact

What I'm looking for

Forward Deployed Engineer, Applied AI Engineer or AI platform architect: close to the customer, responsible for the whole chain from the model to the audit trail. Permanent position or consulting mission, in Paris or remote across EMEA.

The fastest way to evaluate me: pick one project above and ask me how it was built.

oelhachi@gmail.com Download CV
  • What I bringEight years of production engineering, three of them making generative AI work in regulated companies, as the technical lead.
  • Where I'm strongestEnterprise AI platforms and cost governance; agents grounded in documents; AI security; getting teams to adopt what was built.
  • What I don't claimI have not fine-tuned models at scale and I don't publish research. I ship systems and get them used.