LLM-First Math Libraries Engineer & Dev Experience

NVIDIA; France; Sur place; EUR 70 000 - 120 000; Plein temps

Résumé du poste

NVIDIA in France seeks an engineer to accelerate integration of CUDA-X APIs into LLM and agentic AI workflows. You will design LLM-first APIs, collaborate with stakeholders, and ensure high-quality, well-documented code for performance-critical math libraries.

The role emphasizes independent work, strong communication, and the ability to transform complex APIs into easily consumable tools for AI applications across GPU-accelerated platforms.

Qualifications

• PhD or MSc in Computational Science, Computer Science, Applied Math, or related field (or equivalent).
• 3+ years of experience.
• Robust knowledge of LLMs, finetuning, RL, RAGs, MCP, and agenting tooling.
• Proven API design experience with clear, well-documented code in complex domains.
• Ability to prioritize multiple projects and work independently with minimal direction.
• Excellent collaboration, communication, and documentation habits.

Responsabilités

• Drive architecture and implementation of math libraries APIs and docs designed for LLM-first use.
• Collaborate with stakeholders to deliver timely LLM-enhanced library releases.
• Build metrics and tooling to measure impact of LLM/agentic code generation.
• Prototype tooling to transform APIs into LLM-friendly representations for codegen tools.

Connaissances

APIs design

LLM experience

Documentation

Independent work

Collaboration

Formation

PhD or MSc in Computational Science/CS/Applied Math

Outils

CUDA-X APIs

Description du poste

NVIDIA in France seeks an engineer to accelerate integration of CUDA-X APIs into LLM and agentic AI workflows. You will design LLM-first APIs, collaborate with stakeholders, and ensure high-quality, well-documented code for performance-critical math libraries.

The role emphasizes independent work, strong communication, and the ability to transform complex APIs into easily consumable tools for AI applications across GPU-accelerated platforms.