Infrastructure for heterogeneous chips and intelligent systems

Co-founder

Member of Technical Staff - Expression of Interest

Callosum is the Intelligent Systems Company focused on building heterogeneous AI infrastructure. They are looking for exceptional engineers to work on high-performance systems and contribute to the development of novel accelerators and orchestration systems.

Inference Performance & Deployment - Member of Technical Staff

Callosum is the Intelligent Systems Company, focused on co-evolving models, workflows, and silicon to create advanced intelligence solutions. The role involves bridging internal engineering with real-world performance, designing tools and methodologies to ensure production-readiness and systematic performance characterization.

Inference System & Performance - Member of Technical Staff

Callosum is an intelligent systems company building infrastructure that unifies heterogeneous compute across the full stack. The Member of Technical Staff will own end-to-end performance for inference platforms by designing and optimising serving systems, managing caching and scheduling, tuning GPU and network performance, and developing benchmarking and performance-visibility tooling.

Research Engineer, Benchmarking - Member of Technical Staff

Callosum is building infrastructure that unifies heterogeneous compute across the full stack for the next generation of intelligent systems. The Research Engineer will own a reproducible, contamination-controlled benchmarking system for evaluating agentic and algorithmic approaches using execution-based measures of task success, quality, and robustness. The role also leads benchmark co-publications and feeds evaluation results into product, customer, and company-wide quality decisions.

ML Research Engineer - Member of Technical Staff

Callosum is an Intelligent Systems Company building infrastructure that unifies heterogeneous compute across models, workflows, and silicon. The ML Research Engineer will develop evidence-based methods for automatically building and evaluating agentic systems, investigating failure modes, improving intelligence, cost, and reliability, and creating durable research and evaluation infrastructure.

ML Research Scientist - Member of Technical Staff

Callosum is an intelligent systems company building infrastructure that unifies heterogeneous compute across the full stack. The ML Research Scientist will conduct end-to-end research on capable, efficient, and reliable intelligent systems, developing novel approaches, running experiments, and communicating findings with a highly technical team.

Research Engineer, Evals - Member of Technical Staff

Callosum is an Intelligent Systems Company building infrastructure that unifies heterogeneous compute across models and hardware. The Research Engineer will develop rigorous evaluation methodologies, benchmarks, and observability infrastructure for agentic AI systems, transforming experimental traces into evidence that guides system design and predicts capabilities.

Agent Runtime & Systems - Member of Technical Staff

Callosum is building intelligent systems infrastructure that unifies heterogeneous compute across the full stack. The Member of Technical Staff will own the runtime and programming model for long-running agent systems, developing durable execution, distributed scheduling, secure tool execution, observability, and debugging capabilities.

Accelerator Systems Software - Member of Technical Staff

Callosum is the Intelligent Systems company focused on building infrastructure for heterogeneous intelligence. The role involves developing low-level systems software for diverse accelerators, enabling the company to run AI workloads efficiently across various hardware platforms.

Evolutionary Optimisation - Member of Technical Staff

Callosum is an Intelligent Systems Company building infrastructure that unifies heterogeneous compute across the full stack. The Member of Technical Staff will own and advance an LLM-guided program evolution and optimisation loop that generates, evaluates, and improves programs across diverse workloads, while developing reusable infrastructure and benchmarks for increasingly difficult optimisation problems.