> Markdown version of [/jobs/ext/3866963-data-scientist](https://www.wearedevelopers.com/jobs/ext/3866963-data-scientist). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Data Scientist - **Company:** Roche Laboratories Inc. - **Location:** Warsaw, IN, United States (Remote available) - **Salary:** $170,100.0 - $243,000.0 - **Contract:** Permanent contract - **Skills:** Application Programming Interfaces (APIs), Artificial Intelligence, Amazon Web Services, Microsoft Azure, Databases, Continuous Delivery, Python (Programming Language), Machine Learning, Sentiment Analysis, Software Engineering, Systems Architecture, Feature Engineering, Retrieval-Augmented Generation, Large Language Models, Prompt Engineering, Model Validation, Vector Embeddings, Generative AI, Agentic-AI, Low Latency, Production Code, Machine Learning Operations, GXP - **Published:** October 22, 2026 - **Apply:** https://arc.dev/remote-jobs/j/redirect/pthlejsnn5 ## About the Role Technical depth * Hands-on expertise in data science and software engineering. Python, LLMs, RAG pipelines, agentic workflows. Equally comfortable writing production code, designing system architecture, or setting integration standards * Evaluate and fine-tune pre-trained models for domain use. Design RAG and search/synthesis architectures. Build evaluation harnesses and statistically evaluate outputs for safety, reliability, and consistency * Strong NLP foundations: preprocessing, language modelling, sentiment analysis, NER * A solid core in statistics, model evaluation, feature engineering and standard algorithms * Real depth on vector embeddings, search theory, and vector database selection and architecture, along with prompt and context engineering Strategic and operational range * Comfortable moving from strategic roadmap planning to solving ad-hoc technical problems on the ground. Skilled at bridging non-technical stakeholders and technical implementation * Comfortable building MLOps/LLMOps pipelines for continuous deployment and monitoring, including observability for LLM systems (tracing, tokens, latency, cost telemetry and drift and hallucination monitoring). * Comfortable forecasting and optimising infrastructure and API costs, and managing platform vendor relationships to keep tools interoperable, secure, and future-proof Communication * Strong written and verbal English, with the ability to explain technical trade-offs to technical and non-technical audiences alike * A learning posture that keeps pace with a fast-changing field, with enthusiasm for knowledge-sharing. Preferred/nice to have: * Hands-on experience deploying enterprise AI solutions on AWS or Azure (GCP exposure a plus) * Model risk management, fairness and bias mitigation, transparency, explainability, and data privacy * Pharma, healthcare, or financial services, including GxP compliance and handling sensitive information, plus awareness of emerging compliance standards for GenAI and foundation models ## Description Roche's AI Function builds the platforms, tools, and data science muscle that let teams across Roche turn AI, GenAI, and agentic ideas into real products, fast. We make it safe, simple, and scalable to build with LLMs, enterprise data, and orchestration, so experimentation stops staying experimental and starts driving real business value. And now we're pushing further into agentic AI: systems that reason, plan, and act on their own, under human oversight. Join us and help define how Roche builds, evaluates, and governs that next generation of AI at enterprise scale. The kind of work you'll do Build and ship * Design and ship AI/ML solutions, from prototype to a monitored capability people actually depend on, including Roche's next generation of agentic AI: systems that plan and execute complex, multi-step work autonomously. Work closely with product owners, engineers, and business stakeholders throughout * Turn ambiguous problems into clear modelling approaches, and explain the trade-offs in language anyone can follow. Choose the right tool for the problem: a classical model, a fine-tuned model, retrieval over a foundation model or an agentic workflow, rather than reaching for an LLM by default * Write clean, production-ready, reproducible code, and help shape the reusable patterns the wider team relies on Bring rigour * Bring rigour to every experiment. Solid baselines, held-out testing, real error analysis. No shortcuts * Treat evaluation as core craft, not an afterthought. Build eval sets and harnesses for GenAI systems, use golden datasets and regression checks for prompts, and statistically evaluate outputs for safety, reliability and consistency, offline and in production * Root solutions in real user feedback, not just lab results Measure and inform * Power decisions with analytics. Surface the cost, value, adoption, and usage data that tells AI leadership what's working and where to invest next Govern and share * Build within Roche's governance, security, and compliance guardrails, from day one of a project to the last * Contribute to those guardrails in practice: help design the checks that help embed fairness, explainability and privacy into pipelines so they are enforced rather than aspirational * Share what you learn through reviews, demos, and onboarding docs. No lesson stays siloed