Senior Applied Research Engineer
Pubblicato il 25-08-2026 - Jobgether in Italia
This position is listed on behalf of a partner company, who manages all applications and next steps. Our partner is looking for a Senior Applied Research Engineer - Video based in Italy.
As a Senior Applied Research Engineer, you will help build the next generation of production-grade foundation models for human-centric video generation.
You will work at the intersection of generative AI research, large-scale distributed systems, and production engineering.
Your work will focus on developing realistic, controllable, and expressive video generation models that can operate reliably at scale.
You will own research and engineering projects end to end, translating hypotheses and experiments into measurable product impact.
The role combines advanced modeling, distributed training, evaluation, inference optimization, and rigorous experimentation.
You will operate in a highly technical, high-ownership environment where research is expected to move quickly toward real-world deployment.
Your contributions will directly influence AI-powered video products used by businesses around the world.
n
Accountabilities
Develop and scale latent video diffusion models designed for human-centric video generation.
Design advanced conditioning mechanisms that improve control over elements such as pose, emotion, scripts, and camera movement while maintaining high visual fidelity.
Lead end-to-end applied research and engineering projects, from developing hypotheses and running experiments through to production implementation and measurable impact.
Develop and optimize distributed training strategies using technologies such as DDP, FSDP, DeepSpeed, and sequence parallelism.
Improve training stability and efficiency across large-scale, multi-GPU and multi-node environments while working within real-world compute constraints.
Design robust evaluation frameworks combining automated metrics with structured human evaluation to assess model quality and performance.
Optimize model inference for low