Role summary:

You will design, build and evaluate the language-model capabilities behind Energy.AI, from retrieval pipelines over technical documents to agents that reason over structured subsurface data. You will own model quality: how well the system retrieves, answers and acts, and how we prove it.

Key responsibilities:

- Build and optimise RAG pipelines over unstructured content (reports, PDFs, scanned documents) and structured data (databases, well and production data).

- Design chunking, embedding, hybrid search and re-ranking strategies, and measure their impact.

- Develop agentic workflows that plan, call tools and return grounded, cited answers.

- Create and maintain evaluation datasets and metrics (faithfulness, relevance, answer accuracy, latency, cost).

- Apply OCR, document parsing and NLP techniques to extract information from technical documents.

- Experiment with prompt engineering and, where justified, fine-tuning or smaller specialised models.

- Work with domain experts to validate outputs and capture feedback for improvement.

- Package models and pipelines for production with the engineering and DevOps teams.

Required experience and skills:

- 5+ years in data science or ML engineering, including at least 2 years working with LLMs.

- Hands-on delivery of RAG pipelines and LLM evaluation in a production or near-production setting.

- Strong Python and the scientific stack (pandas, NumPy, scikit-learn) plus PyTorch or TensorFlow.

- Experience with LangChain, LangGraph or LlamaIndex, and with vector databases (e.g. Azure AI Search, Qdrant).

- Solid grounding in NLP, information retrieval and statistics.

- Ability to explain model behaviour and trade-offs to non-technical stakeholders.

Preferred / nice to have:

- Experience with Azure OpenAI and Azure Machine Learning.

- Document AI experience (OCR, layout parsing, table extraction).

- Knowledge graphs or text-to-SQL experience.

- Exposure to geoscience, engineering or other technical-document domains.

Technology environment:

Python, PyTorch, LangChain / LangGraph, LlamaIndex, Azure OpenAI, Azure AI Search, Qdrant, Databricks, MLflow, FastAPI.