Callosum

Callosum builds AI infrastructure that distributes workloads across different models and chips. Its Tailored Inference APIs let developers use task-specific combinations of compute without rewriting their applications. The platform breaks work into smaller tasks and assigns each to suitable models and hardware under cost, energy and speed constraints. Production applications span cybersecurity, finance and enterprise automation, with silicon partnerships including Cerebras and Rebellions.

Co-founders Danyal Akarca, CEO, and Jascha Achterberg, CTO, met while completing computational neuroscience PhDs at Cambridge. Their research examined how specialised brain circuits combine to produce intelligence. Callosum’s Principle of Maximum Heterogeneity develops a mathematical framework for studying how specialised components cooperate under resource constraints. Its engineering approach translates workloads into reusable functional blocks, combining deterministic operations with model calls and learning how to route computation across available configurations.

In 2026, Callosum raised a $100 million seed round led by Atomico, with participation from Plural, DCVC and the UK Sovereign AI Fund. Its partner case studies illustrate practical applications: the company reported that Round Treasury’s mean document-extraction time fell from 15.9 to 2.1 seconds per document using Callosum. Round Treasury uses this pipeline to extract structured data from invoices and receipts for finance workflows.