Improbable
Building virtual worlds at massive scale from London
What they look for (Software & Engineering): Improbable looks for software engineers who are comfortable working at the intersection of distributed systems, simulation and real-time networking. Candidates should bring deep technical curiosity, an appetite for solving problems that don't yet have textbook solutions, and the ability to collaborate across disciplines ranging from game technology to defence modelling.
What experience could you bring to building systems that distribute simulation workloads across many machines?
Simulating Worlds That Don't Exist Yet
Improbable was founded in 2012 by Herman Narula and Rob Whitehead, two Cambridge graduates who believed that the way we model and interact with complex systems was fundamentally limited by single-machine thinking. Their thesis was deceptively simple: if you could distribute a simulation across hundreds or thousands of machines, coordinating them so tightly that the result felt seamless, you could build virtual worlds of a scale and fidelity that had never been possible before. More than a decade later, Improbable has grown into one of London's most distinctive technology companies, occupying a space that defies easy categorisation.
Headquartered in the city's West End, Improbable operates at the boundary of several fields: distributed computing, game development, defence simulation and what the company broadly calls the "metaverse." Its core product, SpatialOS, is a platform that allows developers to build and run simulations that span far more entities, interactions and geographic area than a single server could ever handle. The technology partitions a simulated world into manageable regions, each processed by a separate worker, while a central orchestration layer ensures that entities moving between regions experience no visible seam.
From Gaming to Defence and Beyond
Improbable first gained attention in the gaming industry. SpatialOS promised MMO-scale experiences without the brutal engineering compromises that had traditionally kept online game worlds small or sparsely interactive. Studios such as Bossa Studios and Midwinter Entertainment used the platform to prototype ambitious multiplayer titles. Improbable itself established internal game studios, including one in Edmonton, Canada, exploring what gameplay looks like when thousands of players and AI agents coexist in a single, persistent environment.
The company's trajectory shifted materially around 2020 when it began applying its simulation technology to defence and national security. Through its subsidiary Improbable Defence, the company now works with the UK Ministry of Defence, NATO allies and other government clients to create synthetic environments for military planning, training and decision support. These are not simple war games. They are large-scale digital twins of real-world theatres, populated with realistic agent behaviours, logistics chains and environmental conditions, all running in real time. The pivot brought new contracts, new credibility in Whitehall, and a fundamentally different set of engineering challenges.
A Technical Culture Built on Hard Problems
What makes Improbable distinctive as an employer is the sheer difficulty of the problems it tackles. Distributed systems are notoriously hard to reason about: partial failures, network partitions, clock drift and state consistency are daily concerns. Layer on top of that the requirement for real-time performance, where a simulation tick might need to complete in 50 milliseconds across dozens of machines, and you have an engineering environment that is genuinely unusual in the UK tech landscape.
"We're not building another SaaS dashboard. We're trying to make machines collaborate at a speed and scale that lets you forget they're separate machines at all."
The engineering organisation spans several disciplines. Infrastructure teams work on container orchestration, networking and cloud resource management, primarily on Google Cloud Platform and AWS. Platform engineers maintain and extend the core SpatialOS runtime, dealing with entity serialisation, load balancing across workers, and the subtleties of interest management, deciding which workers need to know about which entities at any given moment. Client-facing SDK teams build tooling for external and internal developers in C++, C#, Rust and Go. And an increasingly important cohort of simulation engineers works closely with defence analysts and domain experts to translate real-world complexity into models that behave plausibly under stress.
Culture and Working Life
Improbable's London office, near Spitalfields, reflects the company's stage. It is neither a scrappy startup loft nor a sterile corporate campus. The atmosphere is technically serious, with whiteboards dense in state diagrams and latency budgets, but the dress code is relaxed and the hierarchy relatively flat. Engineers are encouraged to write technical RFCs and to challenge architectural decisions regardless of seniority.
The company has weathered the volatility common to deep-tech ventures. Significant funding rounds, including a $502 million Series B led by SoftBank in 2017, set high expectations. Subsequent years brought restructuring, studio closures and a sharpened focus on defence revenue. Employees who have stayed through these cycles tend to share a pragmatic resilience and a genuine interest in the core technology rather than hype.
Where Improbable Is Heading
The near-term roadmap centres on expanding its defence and government business while continuing to mature its simulation platform for commercial use cases such as urban planning, logistics modelling and telecommunications network design. There is growing interest in how large language models and other AI techniques can populate simulations with more believable agents, reducing the manual effort of scenario authoring. Improbable has begun investing in this intersection, exploring how generative AI can work alongside its distributed infrastructure.
For engineers considering a move, Improbable offers something that is genuinely rare in the London market: the chance to work on infrastructure problems that are closer to academic research than to typical product engineering, but with the urgency and accountability of real customers, some of whom operate in environments where simulation accuracy has life-or-death consequences. It is not a company for everyone. The problems are open-ended, the technology stack is deep, and the business context can shift. But for those who find satisfaction in making complex systems work at the edge of what is technically possible, it remains one of the most compelling places to be in the UK.