Staff Engineer, AI Inference Performance & Deployment

Uncover; Greater London; Hybrid; GBP 110,000 - 160,000; Full time

Benefits offered by this job

Competitive salary

Equity & ownership

Private healthcare

Visa sponsorship and relocation

London office on-site

Inclusive workplace

Job summary

Callosum, based in London, is building infrastructure that unifies heterogeneous compute across the full stack. This role bridges Callosum's internal engineering with real-world performance, sitting at the integration point of every engineering function to ensure production-readiness and empirically guided decisions.

You will run experiments, benchmark performance, and shape deployment patterns across diverse hardware, aiming for scalable, cost-effective inference in production-like environments.

Qualifications

• Experience deploying large model inference in production or production-equivalent environments.

Responsibilities

• Run experiments self-hosting models on cloud or on-prem across providers and hardware configurations, characterising performance envelopes.
• Develop and maintain deployment patterns that are reproducible and optimised for latency, throughput, and cost.
• Work at the orchestration and routing software to improve caching, request scheduling, batching, and resource allocation.
• Serve as integration point for accelerator support and infrastructure upgrades, providing feedback on bottlenecks and capabilities.
• Build and maintain benchmarking harnesses, regression suites, and performance dashboards.

Skills

Production-grade inference

Multi-node GPU deployments

Performance characterization

Clear communication

Tools

Dynamo

Triton Inference Server

Job description

Callosum, based in London, is building infrastructure that unifies heterogeneous compute across the full stack. This role bridges Callosum's internal engineering with real-world performance, sitting at the integration point of every engineering function to ensure production-readiness and empirically guided decisions.

You will run experiments, benchmark performance, and shape deployment patterns across diverse hardware, aiming for scalable, cost-effective inference in production-like environments.