Member of Technical Staff (Research)

Radical Numerics is building the Automated AI Research Factory: a new kind of research lab where AI develops AI alongside human experts. Our first system, RISA-1 (Recursive Intelligence System Architecture) orchestrates environments where agents and humans collaborate to accelerate every part of AI development: from infrastructure and numerics to models and scientific applications. The goal is to unlock recursive self-improvement (RSI) in a collaborative, engineered way—so new architectures, kernels, and models can be designed, tested, and deployed faster than ever. As a Member of Technical Staff, you'll help us push the boundaries of how AI can improve itself. Depending on your background, you might contribute across different areas. AI Systems & Infrastructure - Build environments where agents can run large-scale experiments, from pretraining to inference. - Develop observability and debugging tools to manage swarms of parallel agents. AI Research - Design and prototype new architectures (multi-hybrids, diffusion language models, scientific models). - Automate research protocols that encode expert intuition for agents to execute. Numerics & Hardware-Aware Algorithms - Implement and optimize custom operators and kernels (e.g., Phalanx). - Translate algorithmic ideas into performant code across diverse hardware. Agents & Protocols - Build agent workflows that integrate tool use, memory, and experimental protocols. - Design planning strategies for parallelized and long-horizon research tasks. Domain Applications - Apply RISA to scientific problems such as DNA modeling, sensor data, and beyond. Software engineering - Design and implement robust, scalable systems. - Develop features end-to-end: from architecture and design to testing and deployment.

Radical Numerics is building the Automated AI Research Factory: a new kind of research lab where AI develops AI alongside human experts. Our first system, RISA-1 (Recursive Intelligence System Architecture) orchestrates environments where agents and humans collaborate to accelerate every part of AI development: from infrastructure and numerics to models and scientific applications. The goal is to unlock recursive self-improvement (RSI) in a collaborative, engineered way—so new architectures, kernels, and models can be designed, tested, and deployed faster than ever.