Experienced AI Engineer (Full remote France)

Project overview

Context

A European player in the media and video content sector is looking for an AI engineer to strengthen its AI R&D team. The person will design, evaluate, and deploy AI systems applied to a large catalog of multilingual video content. The position requires moving quickly from prototype to measurement, then to production, within a demanding European framework: AI Act, GDPR, preference for open source and sovereign hosting.

Missions

Technical Stack

Working Conditions

• Intelligent search: build and improve hybrid search engines (lexical and vector) on a multilingual catalog, with query understanding by LLM, towards a unified, multimodal, and agentic search API.
• Multimodal enrichment: design pipelines that automatically describe videos from images, transcription, and metadata, using open-source models run locally on GPUs (vision-language, speech recognition, entity extraction), across an entire catalog.
• AI Agents: develop agents that use internal tools and databases via MCP servers, behind a multi-model gateway, with tracing and cost control.
• Decision support: produce factual dossiers cross-referencing internal data and external signals (search trends, press, social networks, platforms): indicators calculated by code, text written by a model then verified digit by digit.
• Evaluation: maintain ground truth datasets, search metrics, LLM judge, A/A and A/B tests, and monitor latency, tokens, and costs, to quickly compare models and frameworks.
• Production and product: deploy on Kubernetes (CI/CD, secrets, API quotas); frame with product and business teams, document, give demonstrations.
• Languages: Python (main), TypeScript, SQL
• API: FastAPI, Pydantic, asynchronous programming
• LLM: Commercial APIs (Gemini, Claude, Mistral) and open-source models (Qwen, Gemma, Llama); LiteLLM; structured outputs
• Multimodal: Qwen-VL, Gemma, Whisper, CLIP, SigLIP, GLiNER, FFmpeg
• Inference: GPU, vLLM, llama.cpp, quantization
• Agents: Google ADK, LangGraph or PydanticAI, MCP
• Search and data: PostgreSQL/pgvector, OpenSearch or Elasticsearch, Redis
• Evaluation: pytest, ranx, LLM judge, Langfuse
• Front: Next.js, React
• Infrastructure: Docker, Kubernetes, Helm, GitHub Actions, public and European cloud
• Tools: uv, ruff, AI code assistants
• Experience: 3 to 6 years, including at least one year with LLMs in production

Project details

• Start: flexible
• Remote: remote
• Location: FR
• Rate: 63-63 EUR/h
• Languages: French