Applied AI Engineer F/H/X

TEAMWILL; Courbevoie; Sur place; EUR 75 000 - 120 000; Plein temps

Résumé du poste

Teamwill Consulting in Courbevoie, Île-de-France, is seeking an Applied AI Engineer to design, build, and operate production AI systems that integrate LLMs with self-hosted or managed models within strict infrastructure constraints. You will convert prototypes into production-ready services with observability and safety.

Work involves collaboration with Product and Engineering teams in a regulated financial-services environment, with on-site engagement typical three days per week and

Qualifications

• Master’s degree or equivalent in CS, math, or a related engineering discipline.
• 3+ years of hands‑on software engineering experience, including production systems with ML/LLM integration.
• Strong Python skills and at least one additional language (e.g., TypeScript or Java) with API model integration experience.
• Experience with event‑driven architecture and message brokers (e.g., Kafka, RabbitMQ).
• Familiarity with containerization (Docker) and automated test suites.
• CI/CD awareness and Everything as Code mindset (infrastructure as code).
• Comfort with data‑governance constraints and responsible data handling.
• Professional fluency in English; French for France-based roles.

Responsabilités

• Design, build, test, and operate agentic AI systems and pipelines calling one or more LLMs.
• Start with simple compositions, then evolve to autonomous, tool-using agents.
• Integrate managed and self-hosted models into the same application and manage GPU capacity.
• Build and maintain retrieval pipelines ground models in knowledge bases (RAG).
• Develop eval suites to gate changes before production with guardrails and observability.
• Package AI components as containerized services with documented interfaces.
• Use AI coding agents and maintain project context for agents' workflow.
• Collaborate with clients and engineering teams to move assets from prototype to production.

Connaissances

Software engineering

Python

Other language (TypeScript/Java)

Event-driven architecture

CI/CD

Data governance

English fluency

French language

Formation

Master’s degree / equivalent

Outils

Docker

Kubernetes

GitOps

Description du poste

About Teamwill Consulting Putting domain expertise and artificial intelligence at the heart of the decisions that finance millions of projects, vehicles, and equipment around the world is the core of what we do.

About Teamwill Consulting Putting domain expertise and artificial intelligence at the heart of the decisions that finance millions of projects, vehicles, and equipment around the world is the core of what we do. For more than 20 years, Teamwill Consulting has been supporting leading players in banking, consumer finance, and mobility through their strategic and technological transformations. Operating in 11 countries and bringing together more than 800 professionals, the Group combines deep industry expertise, close client relationships, and the ability to deliver on a global scale. Artificial intelligence is a cornerstone of our growth strategy. To carry that ambition, we are strengthening our technology, engineering, and data teams.

Role Summary

Teamwill Consulting’s CTAIO function builds and operates the firm’s reusable artificial intelligence (AI) engineering assets, then deploys them into some of the firm’s most demanding engagements. As an Applied AI Engineer, you design, build, and operate production AI systems that combine managed large language models (LLM) with managed or self-hosted, proprietary or open-weight models, under real infrastructure constraints. You turn an AI capability from a working prototype into something an engineering organization can trust in production: tested, observable, and safe to change.

Key Responsibilities

• Design, build, test, and operate agentic systems and pipelines that call one or more LLMs to complete multi-step tasks. Contribute to clarifying requirements with Product Owners, Product Managers, Business Analysts, and Subject Matter Experts, and to formalizing non-functional requirements (e.g., application reliability, security, performance…).
• Start with the simplest composition that works (typically applying the KISS (keep it simple and stupid) software craftsmanship principle): a direct model call or a short chain, moving to a more autonomous, tool-using agent design once a simpler one is shown to fall short. Favor a lightweight, provider-agnostic integration approach over a heavyweight orchestration framework whenever possible.
• Integrate managed, application programming interface (API)-based models and self-hosted, proprietary and open-weight models into the same application, including the operational work specific to self-hosting: inference optimization and request scheduling under constrained graphics processing unit (GPU) capacity.
• Build and maintain the retrieval pipelines that ground models via retrieval-augmented generation (RAG) in the client’s or firm’s knowledge bases: turning source documents into a well-structured, versioned, and access-controlled form a model can retrieve from reliably.
• Build and maintain evaluation (“eval”) suites that gate every change before it reaches production (capability, safety, regression, and cost/latency thresholds). Implement guardrails (secret redaction, content-safety filtering, and fallback behavior), and instrument every AI-powered service for observability (token usage, latency, tool-call errors, and output quality) so a production issue can be diagnosed from data.
• Package AI components as containerized, deployable services with documented interfaces (API, command-line interface (CLI), Model Context Protocol (MCP)), so they can be integrated into a client’s systems.
• Use AI coding agents (for example Claude Code, Cursor, or OpenCode) as part of your own engineering workflow: maintain the standing context documentation a project keeps for its agents, and review agent-produced code with the same combination of automated checks and human review as code anyone else on the team writes
• Work closely with clients, Data Engineers, Platform Engineers, and Forward Deployed Engineers to move an asset from prototype to a client-ready, secure, and monitored deployment, and contribute field learnings back into Teamwill’s shared asset portfolio, so a solution built for one engagement strengthens every future one.

What You’ll Work On

Your primary engagement sits inside a large, regulated financial-services environment where AI adoption is real and accelerating. Some workloads run against managed, proprietary models under a strict data-governance agreement; others run entirely on infrastructure the client controls, using open-weight models, to meet data-residency and regulatory requirements. Both paths are rich engineering challenges: the managed path calls for rigorous prompt, cost, and safety engineering, and the self-hosted path for model serving and capacity planning in a market where GPU capacity is scarce. You will have the opportunity to build expertise in both. The work is international: you will collaborate with teams and clients across Teamwill Consulting’s geographies.

What We’re Looking For

• Master’s degree (or equivalent professional experience) in computer science, applied mathematics, or a related engineering discipline.
• 3+ years of hands‑on software engineering experience, including at least [1-2 years] building production systems that integrate large language models or other machine learning models.
• Strong Python skills, with comfort reading and writing in at least one additional language used in your team’s stack (for example TypeScript or Java), and experience integrating with managed or self-hosted model providers through an API: prompt design, cost control, and handling of rate limits and failures.
• Experience with event-driven architecture so that your applications can consume or publish events into message brokers (for example Kafka or RabbitMQ).
• Working knowledge of containerization (Docker) and of building automated test suites (unit, integration, and evaluation-style tests for AI behavior).
• Familiarity with continuous integration / continuous delivery (CI/CD) pipelines (you will be expected to contribute to CI implementation) and with treating infrastructure, evaluation suites, and prompts as versioned, reviewable code: an Everything as Code mindset.
• Comfort operating under strict data-governance constraints (no uncontrolled use of client data, no fully autonomous decisions without a human owner) as a normal part of the job, with the high-agency ownership to identify a gap or a better approach and propose it.
• Professional fluency in English; professional French for roles based in France or Tunisia.

Nice to Have

• Experience with RAG and knowledge-base design (for example, Google’s Open Knowledge Format), or with a model-serving or inference stack for open-weight models (for example vLLM or an equivalent), including the basics of quantization and GPU-aware capacity planning.
• Prior experience in banking, capital markets, or another regulated financial-services sector.
• Familiarity with Kubernetes, GitOps deployment practices, and policy-as-code admission controls (for example Kyverno).

Location & Mobility

A minimum of 3 days per week, on site is expected, at Teamwill Consulting offices or at client engagement locations, with travel varying by assignment. Teamwill fosters an inclusive work environment and ensures equal opportunity in all hiring decisions, regardless of gender, age, national or ethnic origin, disability, sexual orientation, gender identity or expression, beliefs, or any other characteristic protected by applicable law.