Senior AI Engineer — LLM Applications (Patent NLP)

The role

As a Senior AI Engineer for LLM Applications, you will own the LLM and information-retrieval layer of the Digital IP platform. You will turn raw patent documents — claims, descriptions, drawings, file wrappers, and prior art — into structured, queryable, decision-grade signal for inventors, the IP Council, product managers, and executives.

This is a high-leverage individual-contributor role on a small, senior team. You will partner closely with the Architecture Lead, with the IP Council, and with R&D leaders across Ingersoll Rand's business units. You will ship production systems — not prototypes — and your work will be visible at the executive level.
 

What you will do

• Design, build, and operate the LLM-driven services that power patent ingestion, classification (IPC/CPC), claim and entity extraction, semantic search, and prior-art retrieval across the Digital IP platform.
• Own the lifecycle of LLM applications on patent data: retrieval design, prompt and chain engineering, structured-output design, evaluation, and continuous improvement on top of managed LLMs.
• Build retrieval systems over Ingersoll Rand's internal corpus and external patent databases (USPTO, EPO, WIPO), including embedding strategy, vector indexes, and hybrid lexical/semantic search.
• Develop rigorous offline and online evaluation: gold sets co-curated with the IP Council, regression suites, hallucination and citation-faithfulness checks, and human-in-the-loop review workflows.
• Productionise LLM-driven systems end-to-end — retrieval indexes, prompt and chain wiring, serving, monitoring, latency and cost control — on Snowflake (Cortex, Snowpark).
• Translate ambiguous business questions from inventors, patent attorneys, and executives into well-scoped problems and shippable products.
• Contribute to the architectural direction of the Digital IP platform: data contracts, service interfaces, and reusable patterns that scale across business units.
• Raise the team's bar on evaluation rigour, documentation, and reproducibility, and mentor engineers and analysts who interact with the LLM stack.
   

What we are looking for

Required

• 5–8 years of professional software engineering experience, with the last 2+ years focused on shipping LLM-driven applications.
• At least one substantial LLM application that you took to production — RAG, agents, or LLM-driven workflows — and can speak to how you designed retrieval, prompts, evaluation, and post-launch monitoring.
• Strong working knowledge of information retrieval: embeddings, vector search, hybrid lexical+semantic approaches, and chunking strategy.
• Hands-on experience with Snowflake — Cortex, Snowpark, or building data and LLM workloads directly on the warehouse.
• Solid software engineering practice: clean Python, tests, version control, code review, and CI/CD.
• A disciplined approach to evaluation: gold sets, regression suites, hallucination and citation-faithfulness checks, and an instinct for measuring quality before shipping.
• Track record of working directly with non-engineering stakeholders — domain experts, analysts, or executives — and translating their needs into shipped systems.
• Excellent written and spoken English; ability to overlap several hours per day with Central European Time.

Strongly preferred

• Experience with structured information extraction at scale (entities, relations, attributes) from messy, multi-format corpora.
• Experience designing human-in-the-loop and annotation workflows with subject-matter experts.
• Familiarity with LLM application tooling — orchestration frameworks, observability, prompt/version management, cost and latency monitoring in production.

Nice to have

• Direct experience with patent or other long-form legal/technical documents: claim structure, IPC/CPC taxonomies, USPTO/EPO/WIPO data, or PATSTAT.
• Fine-tuning of LLMs or embedding models for domain-specific tasks, where the case for it was clear.
• Classical NLP or machine-learning background (training models, MLOps lifecycle).
• Background in mechanical, electrical, chemical, or industrial engineering — or any other domain that makes patent content feel familiar rather than foreign.
• Experience working in a regulated or IP-sensitive environment.