LLM Engineer

Luminary is a Series B company that helps engineering companies enhance their competitiveness through its Physics AI platform. The LLM Engineer will architect the Physics AI Copilot, focusing on developing intelligent assistants that improve engineering workflows by leveraging large language models. Responsibilities: Develop Agentic AI systems: Design and implement tools for agents to call; build reasoning, planning, and orchestration capabilities that enable the copilot to autonomously execute complex engineering workflows; Design and optimize RAG pipelines: Build retrieval-augmented generation systems over engineering documentation, physics simulation results, and domain knowledge bases; Implement memory and context management: Create persistent conversation memory and context systems that maintain coherent, long-running engineering sessions; Fine-tune and adapt LLMs: Customize foundation models for Physics AI and physics simulation domain expertise through fine-tuning, prompt engineering, and evaluation frameworks; Deploy and scale LLM infrastructure: Build robust, production-grade systems for self-hosting and serving LLMs, optimizing for latency, cost, and reliability; Integrate with Physics AI and physics simulation platform: Connect LLM capabilities with Luminary's Physics AI training/evaluation/inference pipelines, physics simulation solvers, mesh tools, and analytics APIs to enable end-to-end automation; Establish evaluation frameworks: Define metrics and build testing infrastructure to measure copilot quality, accuracy, and user satisfaction; Collaborate cross-functionally: Work closely with Physics AI researchers, platform engineers, and product teams to deliver customer-centric AI experiences Qualifications: Bachelor's degree or higher in Computer Science, Mechanical Engineering, Aerospace Engineering, or related field; 5+ years of experience building production software or ML systems; 2+ years of hands-on experience developing LLM-powered applications; Strong proficiency in Python; Proficiency using coding agents such as Claude Code; Experience with Agent Evals; Deep understanding of LLM architectures, prompting techniques, and their capabilities/limitations; Experience designing tools/functions for agents to call, with planning and reasoning; Experience with multi-agent orchestration and coordination; Hands-on experience with RAG systems and memory/context management, including vector databases, embedding models, chunking strategies, and long-running session handling; Experience building MCP (Model Context Protocol) servers to expose tools and capabilities to external agents; Experience with agent frameworks (e.g., LangChain, LlamaIndex, Google ADK, Autogen, Claude Agent SDK, or custom solutions); Familiarity with Physics AI, CAE, or physics simulation domains a plus; Experience fine-tuning LLMs for domain-specific applications; Hands-on experience self-hosting and serving LLMs in production environments Required Skills: Python, LLM-powered applications, Agent Evals, LLM architectures, Prompting techniques, Multi-agent orchestration, RAG systems, Memory management, Context management, MCP servers, Agent frameworks, Physics AI, Fine-tuning LLMs, Self-hosting LLMs, TypeScript, Go, Kubernetes, GPU infrastructure, Cloud platforms