Skip to content
Job boardOpen

AI Engineer – Agents, Skills, and MCP

Original title: Ingénieur IA – Agents, skills et MCP

LeHibou · Paris, France · Hybrid · Senior

  • Paris, France · Hybrid
  • French

Indexed (yesterday) · refreshed constantly

Budget
€75 / h
Start
flexible
Work mode
Hybrid
Contract
Freelance / B2B
Seniority
Senior

Required skills

Nice to have:TypeScriptMicrosoft Copilot StudioLoRA / QLoRA fine-tuningFinancial Services Experience

Languages: French

Job description

An independent bank is seeking an AI Engineer to build knowledge corpora, agents, and skills for their two-phase AI strategy, focusing on collective learning and industrializing prototypes.

Project overview

Our client in the Banking and Finance sector is looking for an AI Engineer – Agents, Skills, and MCP M/F

Mission Description:

Context

An independent bank has embarked on a two-phase AI strategy. Following employee training and the deployment of Microsoft Copilot and Claude, a second phase aims at collective learning, building knowledge corpora, agents, and skills by domain, as well as transitioning prototypes to processes and products. Governance is shared with the CISO and Compliance. The Claude Team platform is migrating to Claude Enterprise, with SSO/SCIM via Entra ID and audit logs. Deployment will initially target the Data team, then the digital bank and the risk management department.

The Data foundation is based on Microsoft Fabric (Lakehouse, Spark, Power BI). The knowledge corpus, centralized by domain, relies on reference sources: code, production files, and business processes. Agents consult this corpus and use reusable skills compliant with the Agent Skills standard, within Claude Projects, Cowork, or Claude Code, with connections to data and tools via MCP servers. For the Data team, the corpus will be built by reverse-documenting the platform, comprising approximately 30 pipelines and 80 notebooks, with code as the primary reference and attention paid to token costs.

Existing assets include Fabric MCP, an NL2DAX data plugin with RLS, and a Corporate Banking agent nearing production. An internal MCP gateway with API Management and OBO flows is envisioned.

Mission

Build the knowledge corpora, agents, skills, and MCP servers necessary for pillars 2 and 3 of the AI strategy, and industrialize the applications currently prototyped in Claude. The mission aims to structure this capability within the Data team, improving quality, security, and cost control. It also includes carrying out POCs for a Go/No-Go decision in Q1 2027 and transferring skills to a future internal role holder.

Main Responsibilities

Technical Environment

Expected Deliverables

Candidate Profile

Experience

Technical Skills

Personal Skills

Modalities

Essential Skills / Qualities:

Advanced Python, AI Agent Engineering and LLM systems in production, Anthropic API and Claude ecosystem, MCP Servers, RAG and agent orchestration, Microsoft Fabric, Power BI and SQL, Security and industrialization of AI agents

  • Build the Data corpus by reverse-documenting code: inventory, dependencies, generated diagrams, production files, and traceability of assertions back to the code.
  • Apply the Agent Skills standard, promote interoperability between Claude Code, claude.ai, and Copilot, and manage token costs.
  • Design and deploy business agents and skills in Claude Projects, Cowork, and Claude Code, then adapt them for the digital bank and risk management.
  • Develop and operate internal MCPs for Fabric and Power BI, including NL2DAX with RLS, Delubac MCP, and the MCP gateway. Take over the Corporate Banking agent after its production deployment to adapt it by vertical.
  • Industrialize Claude applications using Anthropic APIs (Messages, tool use, Agent SDK), CI/CD, and observability.
  • Implement evaluations: test sets, response quality, token costs, and logging.
  • Secure agents with the CISO and Compliance: identities, least privilege, RLS/OLS, read-only by default, prompt injection defense, and client data isolation.
  • Carry out POCs for candidate use cases, measure their value, and prepare for the Q1 2027 Go/No-Go decision. Evaluate, if relevant, self-hosted open-source agent frameworks.
  • Ensure knowledge transfer through pair programming with the Data team, documentation, participation in AI community training, and handover to the future internal role holder.
  • Generative AI and agents: Claude, Anthropic API (Messages, tool use, Agent SDK), Claude Code, Claude Projects, Cowork, Agent Skills, Microsoft Copilot
  • Agents and integration: MCP servers, OAuth, RAG, vector databases, agent orchestration, API Management, OBO flows
  • Development and industrialization: Python, TypeScript, Git, testing, CI/CD, containers, Azure Functions, Azure Container Apps
  • Data and analytics: Microsoft Fabric (Lakehouse, Spark, SQL endpoint, Data Agent), Power BI, DAX, SQL, NL2DAX
  • Security and quality: Entra ID, Key Vault, RLS, OLS, LLM evaluations, observability, guardrails, logging, and token cost management
  • Production Data corpus compliant with the Agent Skills standard, accompanied by a contribution guide.
  • Operational industrialization chain including CI/CD, evaluation, observability, and cost monitoring.
  • Documented Pillar 3 POCs, with value measurement for the Go/No-Go decision.
  • Handover documentation for the future internal role holder.
  • At least 4 years in software or Data engineering, including at least 1 year on LLM or agent systems in production, with references for delivered achievements.
  • Advanced Python; Git, testing, CI/CD, and containers.
  • Anthropic API, Claude Code, Claude Projects, skills, and Agent Skills standard.
  • MCP servers, agent orchestration, RAG, and vector databases.
  • LLM evaluations, observability, guardrails, and token cost management.
  • Microsoft Fabric, Power BI, DAX, and SQL.
  • Azure Functions or Container Apps, API Management, Key Vault, and Entra ID; agent security in a banking environment.
  • Ability to work in pair programming with the Data team, document achievements, and ensure knowledge transfer.
  • Ability to collaborate with the CISO and Compliance on security and governance topics.
  • Staff augmentation mission, at 1 FTE.
  • Duration of 4 months, renewable; role intended for internalization.
  • Reporting to the Head of Data and working in tandem with the AI Platforms Administrator and the Data team, with reviews involving the CISO and Compliance.
  • A proposal facilitating internalization, in the form of a pre-hire, will be considered.
  • Expected response: anonymized CV and availability, daily rate and renewal conditions, two references for LLM or agent systems delivered in production, and a one-page document presenting the approach to corpus construction and agent industrialization.
  • Selection process: pre-selection based on application, followed by a technical interview with the Head of Data and the team.

Nice to have

TypeScript; Microsoft Copilot Studio; LoRA / QLoRA fine-tuning; financial services experience

Project details

Yes — 2 days/week

  • Start: flexible
  • Remote: hybrid
  • Location: Paris, France, FR
  • Rate: 75-75 EUR/h
  • Languages: French

Indexed by 9am's meta-search. You are a job provider? Claim or manage this listing

Found on collective.work

  1. Job boardNew

    Senior Python Developer

    j-labs· Kraków, Zabłocie, Poland· Hybrid· SENIOR· zł 160–180 / h

    • Python
    • REST API
    • FastAPI
    • +5

Browse all jobs
Sign up free to apply