ML Engineer

<h2>ML Engineer</h2> <h3>Company</h3> <p>Orcrist builds the Orcrist Intelligence Platform (OIP), a Kubernetes-based data intelligence system delivered as SaaS or self-hosted/on-prem (including air-gapped deployments). We combine data processing, ML/AI, and a modern web application to support mission-critical customers across public and private sectors.</p> <h3>Role</h3> <p>Incubate and validate new ML initiatives end-to-end. On Innovation, you’ll build adoption-ready prototype vertical slices spanning data flows, model serving, evaluation, and product integration—then hand off clear artifacts so delivery teams can productize and own them long-term.</p> <h3>What you'll do</h3> <ul> <li>Build ML prototype vertical slices that connect ingest/processing to inference and visible product outcomes (search, insights, UX flows).</li> <li>Create evaluation harnesses and decision artifacts: datasets, baselines, quality/latency/cost metrics, and go/no-go recommendations.</li> <li>Package prototypes for adoption: containerize services, define reproducible deployments, and produce runbooks/checklists.</li> <li>Partner with Research and Data Engineering on dataset curation, annotation loops, experiment tracking, and safe iteration.</li> <li>Make prototypes operationally credible: instrumentation, monitoring, and security/compliance basics (PII handling, provenance mindset).</li> </ul> <h3>About You</h3> <ul> <li>3+ years ML engineering/MLOps experience (level dependent), with evidence of shipping real systems.</li> <li>Strong Python and hands-on PyTorch/Transformers; comfortable taking models from notebook to reproducible services.</li> <li>Practical Kubernetes + containers experience; able to deploy and troubleshoot in production-like clusters (including offline/air-gapped constraints).</li> <li>Strong evaluation discipline and monitoring mindset; comfortable communicating tradeoffs clearly.</li> <li>Eligible to work in Germany; EU/NATO citizenship preferred and export-control screening applies.</li> </ul> <h3>Nice‑to‑haves</h3> <ul> <li>GPU serving/optimization experience (Triton/KServe, ONNX/TensorRT, batching, quantization).</li> <li>Streaming/pipeline tooling (Kafka, Ray, Beam/Flink/Spark) and search/vector/graph integrations.</li> <li>German language (B1+) and/or experience with regulated/public-sector datasets and workflows.</li> </ul> <h3>What We Offer</h3> <ul> <li>Modern ML stack in real constraints: Kubernetes, streaming, and hybrid/on-prem/air-gapped deployments.</li> <li>Remote-first in Germany with regular Berlin workshops, 30 days vacation, equipment &amp; learning budget.</li> <li>High leverage: your prototypes and handoffs unblock multiple delivery teams.</li> </ul>