The Enterprise AI organization at Humana is looking to develop advanced agentic systems capable of autonomously managing and decomposing complex goals to improve care quality and experience for millions of consumers.
Requirements
- Proficiency in SQL, Python, and data analysis/data mining tools.
- Experience with machine learning frameworks like Pytorch, Jax, LangChain, LangGraph or AutoGen
- Experience with high performance, large-scale ML systems
- Experience with language modeling with transformers
- Experience with symbolic planning, causal reasoning, and model-based RL
- Experience with large-scale ETL
- Demonstrated experience shipping autonomous, or semi-autonomous agents in production or simulation
Responsibilities
- Architect goal setting and goal decomposition mechanisms for autonomous agents operating in uncertain or open-ended environments
- Design and implement dynamic planning systems (e.g., hierarchical planning, curriculum learning, scratchpad/self-refinement loops)
- Collaborate on memory, tool-use, and feedback-loop designs enabling multi-step, self-directed agent behavior
- Develop evaluation frameworks for alignment with human intent, consistency, progress against long-horizon objectives
- Prototype agents that interface with APIs, MCP servers, search engines, databases, or real-world actuators to pursue goals safely and efficiency
- Explore mechanisms to detect and mitigate goal misalignment, looping behavior, or undesirable emergent strategies
- Work across teams to integrate goal alignment with safety, alignment, and operational reliability mechanisms
Other
- Master's Degree and 4+ years of experience in research/ML engineering or an applied research scientist position preferably with a focus on developing production-ready AI solutions
- 2+ years of experience leading development of AI/ML systems
- Ph.D. in Computer Science, Data Science, Machine Learning, or a related field
- Occasional travel to Humana's offices for training or meetings may be required.
- Work from a dedicated space lacking ongoing interruptions to protect member PHI / HIPAA information
