Location: Oak Ridge, TN
Cap-Exempt H-1B Position — No lottery required
Requisition Id 16166 **Overview:** The Programming Systems Group at ORNL seeks a forward‑leaning Postdoctoral Researcher to advance research at the nexus of Agentic AI, high‑productivity programming systems (Mojo, Julia, Rust, Python), and HPC system co‑design. This position is embedded within the larger DOE ASCR ecosystem, with direct relevance to ongoing efforts, and related AI‑for‑HPC thrusts that emphasize modernization of scientific programming languages and workflows. This aligns with internal project directions emphasizing high‑productivity languages (Python, Julia, Rust) and their emerging competitors/frameworks (Mojo and JAX) for extreme‑scale heterogeneous systems. The selected researcher will explore how autonomous AI agents and LLM‑driven code generation can co‑evolve with next‑generation language and compiler ecosystems to accelerate scientific software development at scale. **Majot Duties/Responsibilities:** - Agentic AI for High‑Productivity Languages: - Develop multi‑agent reasoning systems that generate, refactor, and validate scientific code written in Mojo, Julia, Rust, and Python. - Integrate AI‑driven workflows into DOE‑relevant HPC toolchains, leveraging insights from projects in LLM‑driven HPC programming. - Incorporate feedback from compiler/runtime systems, including MLIR-based ecosystems (Mojo, Julia, Rust). - Compiler Infrastructure & LLVM/MLIR Integration: - Extend LLVM/MLIR pipelines to support AI‑guided optimizations across languages (Julia/JACC, Mojo/MLIR, Rust/LLVM). - Incorporate Enzyme-based automatic differentiation and multi-language IR tooling for AI‑driven analysis. - High‑Productivity Programming Systems R&D: - Contribute to DOE goals of enabling performance‑portable high‑productivity languages (Python/Julia/Rust) and evaluate the emerging role these languages and frameworks within scientific workloads. - Conduct research on language front‑end abstractions, mixed‑precision modeling, heterogeneous parallelism, and MLIR-level tra