Full Stack- LLM Engineer

Full Stack- LLM Engineer at PranaTree Full Stack- LLM Engineer at PranaTree in Berkeley, Illinois Posted in 1 day ago.

Type:

full-time

Job Description:

Full Stack LLM Engineer Role summary The Full Stack LLM Engineer owns the end-to-end delivery of LLM-powered products, from discovery and solution design through development, deployment, and production optimization. This role combines applied LLM engineering, experimentation, backend and frontend product development, and operational excellence to build reliable user-facing AI systems such as RAG applications, AI agents, and copilots. Key responsibilities Own end-to-end LLM solutions, leading discovery, architecture, technical planning, development, deployment, and iteration for products such as RAG systems, AI agents, and copilots. Translate business needs into scalable technical designs, delivery plans, and implementation roadmaps for production AI systems. Build and optimize LLM inference pipelines, including prompting, tool use, function calling, caching, latency optimization, and cost control. Implement guardrails, grounding strategies, evaluation harnesses, fallback behavior, and production monitoring to improve reliability, observability, and safety. Run rigorous experiments across fine-tuning and PEFT methods, embeddings, chunking, retrieval strategies, reranking, and prompt or model variants. Develop offline and online evaluations for output quality, safety, hallucination reduction, groundedness, and model or prompt drift. Document experimental findings and engineering decisions to guide model selection, prompt iteration, and product improvements. Deliver full-stack product components, including user-facing applications, backend APIs and services, authentication, workflow orchestration, and data integrations. Partner with product, design, data, and platform teams to ship AI features that are usable, scalable, and maintainable in production. Required qualifications Proven experience building and shipping end-to-end AI or LLM-powered applications in production across frontend, backend, and model integration layers.