AI Researcher R&D & High-Performance Inference

About the Role
To drive the development of breakthrough technologies, we are seeking AI Researchers. Selected by the World Summit as a strategic Startup Partner, incubated at Station F (one of the largest startup campuses in the world), and recognized by HEC Paris for its exceptional innovation potential, CALLVIUS develops the critical software orchestration layer enabling retail and fintech enterprise leaders to safely deploy, govern, and scale autonomous AI agents.
Reporting directly to the R&D team, you will design, train, and optimize state-of-the-art artificial intelligence models while driving forward-looking research initiatives.
Operating at the exact intersection of foundational research and production engineering, your mission is to build breakthrough architectures engineered to run at scale on next-generation cloud inference infrastructure.
Key Responsibilities
1. Algorithmic Research & Innovation
- Design and explore novel foundation model architectures (LLMs, generative, and multimodal models).
- Maintain an active scientific watch and benchmark/reproduce state-of-the-art (SOTA) research.
- Document methodologies and showcase innovations through technical reports, patents, or publications.
2. Distributed Training & Advanced Optimization
- Build, scale, and maintain massive distributed training pipelines.
- Apply advanced model optimization techniques (quantization, pruning, distillation, and efficient attention mechanisms).
- Identify and eliminate compute, memory, and bandwidth bottlenecks across training clusters.
3. High-Performance Inference & Cloud Deployment
- Design ultra-low latency, high-throughput inference architectures tailored for real-time production environments.
Education & Experience
- PhD in Artificial Intelligence, Machine Learning, Computer Science, Applied Mathematics, or equivalent proven industrial R&D track record.
- Demonstrated track record in applied research or advanced R&D, backed by publications in top-tier conferences (NeurIPS, ICML, ICLR, CVPR, ACL) or significant open-source contributions.
Technical Skills
- Frameworks & Ecosystem : In-depth mastery of PyTorch (mandatory) and familiarity with JAX or modern distributed training frameworks (DeepSpeed, Megatron-LM, FSDP).
- Inference Engines : Hands-on experience deploying and serving models with modern runtimes (e.g., vLLM, TensorRT-LLM, Triton Inference Server, TGI).
- Software Engineering : Advanced proficiency in Python, object-oriented design, data/code versioning (Git, DVC), and testing/CI-CD best practices.
Soft Skills
- Scientific Curiosity & Rigor : Ability to formulate sound hypotheses, test unconventional solutions, and systematically analyze edge cases.
- Pragmatism : Continuous focus on bridging theoretical model performance with production constraints.
- Communication : Strong ability to collaborate seamlessly with software engineers and distill complex mathematical concepts for cross-functional teams.
Compensation
Aligned with your level of expertise.

Work arrangement
Yes