Principal AI Engineer

Job description

**Working with Us** Challenging. Meaningful. Life-changing. Those aren’t words that are usually associated with a job. But working at Bristol Myers Squibb is anything but usual. Here, uniquely interesting work happens every day, in every department. From optimizing a production line to the latest breakthroughs in cell therapy, this is work that transforms the lives of patients, and the careers of those who do it. You’ll get the chance to grow and thrive through opportunities uncommon in scale and scope, alongside high-achieving teams. Take your career farther than you thought possible. Bristol Myers Squibb recognizes the importance of balance and flexibility in our work environment. We offer a wide variety of competitive benefits, services and programs that provide our employees with the resources to pursue their goals, both at work and in their personal lives. Read more: careers.bms.com/working-with-us . ## **Key Responsibilities** - Design, build, and deploy autonomous multi-agent workflows using orchestration frameworks such as LangGraph, CrewAI, Autogen, or similar, including complex state machines with conditional routing, parallel execution, and error recovery patterns. - Architect graph-based agent workflows with 10+ nodes involving agent collaboration, task decomposition, and sequential/parallel execution across multiple business domains. - Develop and maintain reusable agent node libraries, extensible platform patterns, versioning strategies, and testing frameworks (unit, integration, and end-to-end) for agent workflows. - Build production-grade FastAPI applications with async I/O patterns, integrating PostgreSQL, Redis, and external enterprise services. - Implement real-time agent streaming using Server-Sent Events (SSE) and WebSocket protocols, alongside RESTful and event-driven API architectures for agent orchestration. - Integrate cloud-based LLM providers (AWS Bedrock, Azure OpenAI, Anthropic Claude, OpenAI GPT-4) and design prompt management systems with versioning, templating, and dynamic compilation. - Implement conversation state persistence using Redis checkpointing and build tool-calling protocols (Model Context Protocol, function calling) for external data sources and APIs. - Develop hybrid intelligence patterns combining LLM reasoning with rule-based logic and statistical analysis, and build response transformation pipelines for structured analytical outputs. - Integrate observability platforms (Langfuse, LangSmith, or similar) to enable end-to-end agent tracing, telemetry, performance monitoring, and cost optimization across production workflows. - Implement evaluation frameworks measuring agent success rates, reasoning quality, and output accuracy, while continuously optimizing token usage and LLM costs. - Ensure enterprise security integration (LDAP, SSO, access control), robust error handling, and compliance with data governance and Responsible AI standards. - Partner with data engineers, business analysts, and UX teams to translate requirements into scalable agent workflows and streaming interfaces. - Mentor junior engineers on async Python patterns, agent design, and LLMOps best practices; participate in architecture reviews and contribute to documentation and knowledge sharing. ## **Qualifications & Experience:** - Bachelor's or Master's degree in Computer Science, AI/ML, Engineering, Data Science, or a related discipline. - 9+ years of software engineering experience, with 2+ years building and deploying production LLM-powered applications. - Proven experience with: - Agentic orchestration frameworks: LangGraph (preferred), LangChain, CrewAI, Autogen, or similar. - Cloud LLM providers: AWS Bedrock, Azure OpenAI, Anthropic Claude, or OpenAI GPT-4. - Production async web frameworks, particularly FastAPI. - Docker containerization, Git version control, and CI/CD pipelines. - Cloud platforms: AWS, Azure, or Google Cloud. - Expert-level async Python programming (asyncio, async/await patterns). - Demonstrated experience building autonomous agent systems performing multi-step tasks beyond simple chatbots, including conditional logic, state management, and tool-calling patterns. - Strong understanding of agentic design patterns including ReAct, Plan-and-Execute, and tool-use agents. - Experience implementing Model Context Protocol (MCP) and agent-to-agent (A2A) communication frameworks. - Familiarity with LLMOps practices, observability tooling (Langfuse, LangSmith), and prompt engineering at scale (templating, versioning, optimization). - Excellent analytical, problem-solving, and communication skills with the ability to work effectively in globally distributed teams. - Prior experience in global life sciences, especially in the GPS functional area, is a plus. - Experience managing or collaborating with offshore technical development teams and diverse international stakeholders is a plus. *If you come across a role that intrigues you but doesn’t perfectly line up with your resume, we encourage you to apply anyway. You could be one step away from work that will transform your life and career.* **Uniquely Interesting Work, Life-changing Careers** With a single vision as inspiring as “Transforming patients’ lives through science™ ”, every BMS employee plays an integral role in work that goes far beyond ordinary. Each of us is empowered to apply our individual talents and unique perspectives in a supportive culture, promoting global participation in clinical trials, while our shared values of passion, innovation, urgency, accountability, inclusion and integrity bring out the highest potential of each of our colleagues. **On-site Protocol** BMS has an occupancy structure that determines where an employee is required to conduct their work. This structure includes site-essential, site-by-design, field-based and remote-by-design jobs. The occupancy type that you are assigned is determined by the nature and responsibilities of your role: Site-essent