Full-Stack AI Engineer / Data Scientist – Agentic Systems

At Roche you can show up as yourself, embraced for the unique qualities you bring. Our culture encourages personal expression, open dialogue, and genuine connections,  where you are valued, accepted and respected for who you are, allowing you to thrive both personally and professionally. This is how we aim to prevent, stop and cure diseases and ensure everyone has access to healthcare today and for generations to come. Join Roche, where every voice matters.

###

The Position

About the Role

At Roche Digital Technology, we are advancing the boundaries of Applied AI. The Applied AI Use Case Engineering & Operations Team is tasked with building innovative AI applications, GenAI agents, and agentic foundations.

In the 2026 tech landscape, the lines between traditional disciplines have blurred. We operate in small agile teams (e.g., ~6 members) powered by advanced coding agents (like Claude Code) to develop and ship solutions faster than ever before. We are looking for a highly skilled, hands-on Full-Stack AI Engineer / Data Scientist with a deep sense of ownership. Rather than being a narrowly specialized Data Scientist, ML Engineer, or MLOps Engineer, you will combine these skill sets. You will build agentic generative AI systems and classical ML systems end-to-end, taking accountability from concept and exploratory data analysis all the way to production releases and monitoring.

Core Tech Stack & Scope

This role involves working deeply with multimodal foundation models, advanced agentic workflows (including Agent-to-Agent/A2A communication), Model Context Protocol (MCP), Retrieval-Augmented Generation (RAG) pipelines, other emerging AI technologies, and MLOps subsystems. You will utilize cloud services (AWS and multicloud) alongside vector, graph, and traditional databases to develop scalable and robust AI solutions.

Key Responsibilities

Agentic & GenAI Application Development:

• Design and build advanced AI agentic systems, state machines, search-based conversational systems that solve complex business problems.
• Develop workflows leveraging Large Foundational Multimodal Models to process and reason across text, audio, and video modalities.
• Implement Model Context Protocol (MCP) servers/clients to standardize context exchange between agents, data sources, and external tools.
• Collaborate with AI Architects, Product Owners, and fellow developers to integrate AI capabilities into scalable, fair, and ethical end-user applications focusing on relevance and real-time performance.

Full-Stack Engineering & Agentic SDLC:

• Leverage AI coding agents (e.g., Claude Code) daily to accelerate full-stack development cycles, maintaining high productivity across frontend, backend, and infrastructure tasks.
• Take end-to-end accountability for features: write high-quality, production-ready Python (and occasionally TypeScript) code with comprehensive testing and documentation.
• Manage the DevOps/MLOps lifecycle: containerize applications using Docker, configure CI/CD pipelines, and architect high-throughput, reliable cloud-native solutions on AWS/multicloud.

Data Science, EDA & Strategy:

• Perform thorough Exploratory Data Analysis (EDA) to understand dataset characteristics, uncover patterns, detect biases, and identify data quality issues.
• Use statistical and visualization techniques to inform feature engineering, model selection, and optimization of foundation model-based applications.
• Design robust data pipelines to curate, preprocess, and structure diverse datasets that maximize LLM effectiveness and reduce bias.

Algorithm Development & Optimization:

• Design, customize, optimize, and fine-tune LLM-based and traditional AI algorithms for specific use cases (e.g., text generation, summarization, AI agents, sequence modeling).
• Lead advanced prompt engineering strategies, utilizing zero-shot, few-shot and other paradigms to optimize model outputs without extensive fine-tuning.
• Implement pre-generative AI models (e.g., classification, clustering, regression) when they provide a more efficient, interpretable, or cost-effective solution compared to LLMs.
• Optimize model inference speed, reduce latency (cold start reduction, caching strategies), and manage resource usage across cloud architectures.

Evaluation, Observability & Continuous Improvement:

• Conduct rigorous experimentation (A/B testing) and implement automatic metric pipelines (e.g. BLEU/ROUGE, RAG retri…