Das ist der Job
End-to-end training platform: design and build the onprem ML platform covering the full training lifecycle.
Darum lohnt es sich
Pillar 2: LLMOps (internal AI tooling) On-prem LLM hosting:own the infrastructure that backs our internal AI capabilities Agentic workflow infrastructure: partner with AI workflow engineers to translate requirements into scalable platform capabilities supporting the agentic toolswith on-prem LLMs Model lifecycle management: manage multiple open-weight models, fine-tuning pipelines, and versioned rollouts from experiment to production Your Profile: MLOps experience for edge devices:you built MLtraining pipelines and model registries that teams rely onto build products.
Model optimization & edge deployment:pruning,quantization, and deploymen to edge devices for real-time LiDAR applications. Self-service for MLengineers:build platform capabilities that let ML engineers deploy independently, working cross-functionally with ML, Data, and DevOps.
LLM serving & infrastructure:experience self-hosting and scaling LLMs on-prem, ideally with vLLM or comparable serving frameworks.
You understand throughput, latency tradeoffs, and GPU resource management Platform engineering mindset:you care about developer experience, write documentation others use, and treat reliability and observability as first-class concerns.
Skills we expect: Fluent in English (German is a plus) Ownership mindset: you don't just surface problems, you own the solution. Tech stack you can expect: Languages & Frameworks: Python, C++, CUDA, TensorRT, PyTorch,Tensorflow #J-18808-Ljbffr
