AI Research Engineer

Join Imbue as an AI Research Engineer focused on building large-scale reasoning models that can robustly handle complex coding and theorem-proving tasks. You will push the boundaries of agentic AI by developing novel architectures and training methodologies that prioritize logical consistency and systematic generalization over simple pattern matching. Your work will directly contribute to our mission of creating personal computers that can understand our goals and execute them autonomously.

- Develop and implement novel pre-training and fine-tuning objectives specifically designed to enhance logical reasoning and code synthesis. - Design data pipelines that extract high-reasoning traces from diverse sources including formal proofs and technical documentation. - Experiment with reinforcement learning from human and automated feedback (RLHF/RLAIF) to improve agent reliability. - Optimize distributed training infrastructure to scale model sizes and context windows efficiently across large GPU clusters. - Collaborate with the product team to evaluate agent performance on real-world coding benchmarks and iterative debugging loops. - Stay at the forefront of LLM research to rapidly prototype and integrate new architectural improvements.

- Proven track record of training large-scale transformer models (LLMs) or complex reinforcement learning agents. - Deep expertise in PyTorch and distributed training frameworks like Megatron-LM or DeepSpeed. - Strong background in programming languages, compilers, or formal verification. - Experience implementing and scaling novel reasoning techniques such as Chain-of-Thought, STaR, or tree-of-thought search. - Ability to write high-performance C++ or Rust for custom CUDA kernels or data processing pipelines. - Self-directed research style with a history of shipping high-impact open-source projects or peer-reviewed publications.