NLP Engineer

What You Can Expect

Position Summary

Zimmer Biomet runs on text. Regulatory submissions, clinical evaluation reports, complaint records, quality documentation, surgical technique guides, contracts, and support tickets all hold information that people currently find by reading. The Enterprise AI team builds the systems that read it for them.

The NLP Engineer designs, builds, and ships those systems. You will work across the full range of the discipline: retrieval and generation pipelines built on large language models, structured extraction from long and messy documents, and the classical modelling and evaluation work that tells you whether any of it is actually correct. This is a delivery role. You will own use cases from problem framing through to a service running in production with measured quality.

Scope and Impact

You will take business problems posed in natural language and turn them into working systems. A quality engineer who spends two days a month collating complaint themes, a regulatory writer reconciling a submission against source evidence, a commercial team searching thousands of technique guides: each is a use case you will scope, prototype, evaluate, and deliver.

The work is high leverage and highly visible. Enterprise AI serves every function and region at Zimmer Biomet, so a pipeline you build for one business unit is frequently generalised across several. Because we operate in a regulated industry, correctness matters more than novelty. You will be expected to prove that what you build works, to know where it fails, and to say so clearly.

Mode of Work: 2 days WFH,3 WFO
Location: Bangalore

How You'll Create Impact

Key Responsibilities

LLM and Generative AI Engineering

• Design and build retrieval augmented generation pipelines, including chunking strategy, embedding selection, vector storage, retrieval tuning, and reranking.
• Develop and maintain prompt and context engineering for production use cases, treating prompts as versioned, tested artefacts rather than ad hoc text.
• Build agent and tool-calling workflows where a single model call is insufficient, including task decomposition, tool integration, and failure handling.
• Build evaluation harnesses for generative systems, covering groundedness, factual accuracy, hallucination rate, refusal behaviour, latency, and cost per request.
• Work with Azure OpenAI and other approved model providers, selecting models appropriate to each task and balancing quality against cost and latency.

Document and Text Extraction

• Build extraction pipelines that turn long, unstructured, and inconsistently formatted documents into reliable structured data.
• Implement named entity recognition, classification, summarisation, and relation extraction against domain-specific text such as clinical, regulatory, quality, and commercial content.
• Handle real-world document conditions including scanned PDFs, OCR output, tables, multi-column layouts, and inconsistent templates.
• Define and measure extraction quality with precision, recall, and error analysis on a held-out set, and drive quality up through targeted iteration rather than guesswork.
• Build human-in-the-loop review steps where the confidence threshold or the regulatory context requires them.

Modelling and Evaluation

• Select, fine-tune, and evaluate language models and embedding models against clearly defined success criteria for each use case.
• Build and maintain semantic search and similarity capability, including embedding strategy and index management.
• Create and curate labelled datasets, including annotation guidelines and inter-annotator agreement where the task warrants it.
• Establish baselines before reaching for complexity, and demonstrate that added complexity earns its place.
• Conduct error analysis on production systems and feed findings back into the pipeline, the prompt, the model choice, or the data.

Engineering and Delivery

• Write production-quality Python: tested, reviewed, documented, and maintainable by someone other than you.
• Deploy NLP services on Azure and support them in production, working with platform engineers on packaging, CI/CD, and observability.
• Monitor deployed systems for quality drift, cost, and latency, and act on what the monitoring shows.
• Handle sensitive data correctly, applying Zimmer Biomet policies for data privacy, information security, and responsible AI, and flagging use cases that may carry regulatory or validation obligations.
• Document design decisions, evaluation results, and known limitations so that the work can be reviewed, audited, and built on.
• Collaborate with business stakeholders to frame problems, agree success criteria, and manage expectations on what is and is not achievable.

What Makes You Stand Out

Required Technical Skills

• Strong Python, including the scientific and NLP ecosystem such as PyTorch or TensorFlow, Hugging Face Transformers, spaCy, scikit-learn, and pandas.
• Practical command of LLM application frameworks and patterns, for example LangChain, LlamaIndex, or equivalent.
• Experience with vector databases and embedding-based retrieval, for example Azure AI Search, FAISS, pgvector, or Pinecone.
• Solid understanding of transformer architectures, tokenisation, embeddings, fine-tuning approaches, and their practical trade-offs.
• SQL and comfort working with data at scale.
• Software engineering fundamentals: version control with Git, automated testing, code review, and CI/CD.
• Cloud experience, ideally Microsoft Azure, including deploying and running a service rather than only notebook-based development.

Attributes

• Rigorous about measurement, and willing to report a negative result rather than a flattering demo.
• Pragmatic about complexity, favouring the simplest approach that meets the requirement.
• Excellent written and verbal English communication, with the ability to explain model behaviour and its limits to a non-technical audience.
• Able to work independently across time zones with distributed stakeholders and to manage several use cases in parallel.

Preferred

• Experience in a regulated industry such as medical devices, pharmaceuticals, or healthcare, and familiarity with the documentation expectations that come with it.
• Experience with clinical, regulatory, or quality text, or with standards and terminologies such as SNOMED, MedDRA, or ICD.
• Experience with document AI and OCR tooling such as Azure Document Intelligence.
• Experience with MLOps practice including model registries, experiment tracking, and automated retraining or re-evaluation.
• Familiarity with responsible AI practice, including bias evaluation, model cards, and AI governance frameworks.
• Contributions to open source NLP tooling, publications, or comparable public technical work.
• Experience in a Global Capability Center supporting global business functions.

Your Background

Qualifications

Education

• Bachelor’s or Master’s degree in computer science, computational linguistics, data science, engineering, mathematics, or a related field, or equivalent practical experience.

Required Experience

• Three to five years building natural language processing or machine learning systems, including at least one that reached production and real users.
• Demonstrable experience building applications on large language models, including retrieval augmented generation and prompt engineering for production workloads.
• Experience extracting structured information from unstructured documents at scale.
• Experience defining evaluation criteria and measuring model or pipeline quality quantitatively, not by inspection alone.
• Experience working directly with non-technical stakeholders to turn an ambiguous business problem into a defined technical scope.

Working Relationships

This role reports to the Director of AI within the Enterprise AI function. It works closely with AI platform and data engineers on infrastructure and deployment, with information security and privacy teams on data handling, with quality and regulatory colleagues on validation expectations, and directly with business stakeholders across all regions who own the use cases.

Physical Requirements

Travel Expectations