Industrial AI Application Engineer
About the role
Work on site with our customers' engineering and operations teams to turn cutting-edge AI into working solutions on the factory floor. You will own each use case end to end, from the operational problem and raw plant data to deploying and scaling ML and agentic AI on-premises or in the cloud.
You'll make a difference by
- Build data and knowledge pipelines: Ingest, clean and contextualise sensor and time-series data and unstructured documents for analytics, model training and retrieval.
- Build agentic and LLM applications: Develop AI agents with planning, memory, tool use and RAG, including long-horizon agents for tasks such as root-cause analysis.
- Architect end-to-end AI systems: Design scalable, secure stacks spanning databases, microservices, APIs and user applications.
- Ensure quality and reliability: Define evaluation metrics, guardrails and monitoring for models and agents in safety-critical settings.
- Operationalise AI: Run MLOps/LLMOps pipelines, monitor drift, manage retraining, and optimise latency, cost and accuracy across edge, on-premises and cloud.
- Size and serve LLMs: Serve LLMs across multiple GPUs and servers, and translate customer requirements into accurate GPU and server sizing.
- Be the customer's trusted AI expert: Turn complex AI into practical recommendations for engineering, operations and business stakeholders.
Your defining qualities
- Proven industrial AI deployment: 5+ years in Data Science, AI or ML, with confirmed hands-on experience deploying AI into live operation in plants, fabs or facilities. You can walk us through specific deployments, your role and the results.
- Data preparation: Cleaning, validating and feature engineering for noisy, incomplete real-world data, including time-series and sensor data.
- AI agents for data cleaning: Proven experience building agents that automate data profiling and cleaning (missing values, outliers, unit and timestamp issues, tag-to-asset mapping), with validation checks and human review.
- Model training and tuning: Cross-validation, loss selection, regularisation and hyperparameter tuning (grid, random, Bayesian).
- Mathematical and research-driven thinking: Strong grasp of the maths behind ML and neural networks; able to frame industrial problems such as root-cause analysis mathematically, implement algorithms from research papers and prove feasibility through controlled experiments.
- Production LLM serving and sizing: Multi-GPU serving, continuous batching, KV-cache management and quantisation; able to estimate GPU memory, throughput and cost, and load-test against latency targets.
- Industrial domain exposure: Production, process, equipment or facility data in semiconductor, pharmaceutical, chemical, FMCG or other advanced manufacturing.
- Self-starter with engineering rigour: Curious and accountable; verifies every part of a solution, including AI-assisted code, and explains trade-offs clearly to technical and business audiences.
- Education: Degree in Data Science, Computer Science, AI, Engineering, Automation or a related field.
Technologies you should be familiar with
- Core AI/ML: Python, PyTorch, NumPy, pandas, scikit-learn, transformer architectures
- Time-series AI: Time-series foundation models (e.g. Chronos, TimesFM, Moirai), statsmodels, sktime, Darts
- Agents and retrieval: LangGraph, LangChain, RAG, vector databases, function calling, MCP
- LLM serving: vLLM, SGLang, TensorRT-LLM, Triton Inference Server / NVIDIA NIM, Ray
- LLM adaptation: Hugging Face Transformers, LoRA/QLoRA, quantisation (FP8, AWQ, GPTQ)
- Data and backend: PostgreSQL, MongoDB, Neo4j, Kafka, FastAPI
- Deployment and MLOps: Docker, Kubernetes (NVIDIA GPU Operator), Azure / AWS / Google Cloud, MLflow, LangSmith
- Industrial data: Time-series databases, OPC UA, MQTT
Nice to have
- Unstructured data and RAG: Document parsing, OCR, chunking, embeddings and retrieval evaluation.
- Time-series AI: Forecasting, anomaly detection and drift handling, plus how time-series foundation models work and when to use them.
- Advanced agentic AI: Multi-agent orchestration, long-horizon agents (state persistence, error recovery, human-in-the-loop) and agent evaluation.
- Software and system design: Production-grade Python, CI/CD, and full-stack design with microservices, databases and event-driven integration.
- Advanced LLM training: SFT, RLHF/DPO, or adapting LLMs to industrial domains.
- Predictive maintenance or computer vision for visual inspection.
- OT and AI cybersecurity: IEC 62443, network segmentation, secure AI deployment and LLM risks such as prompt injection; GICSP, GRID or ISA/IEC 62443 certification is a plus.
- Large-scale distributed computing: Multi-node inference or training on GPU clusters (NVLink, InfiniBand), or edge deployment.
Public Notice: Recruitment Fraud
We wish to caution jobseekers that Siemens, as a policy, does not authorize external parties/agents to conduct employment drives or extend Offers of Employment on our behalf. We do not ask for bank details, payment of any kind nor personal financial information in return for assurance of employment. Please visit our job portal for all Siemens job openings and apply via the portal.

