GenAI Platform Engineer — Open-Weights & GPU Serving

Deliveroo; Greater London; On-site; GBP 90,000 - 130,000; Full time

Job summary

Deliveroo is seeking a Software Engineer for the GenAI Platform to build production infrastructure for Generative AI. You will work on open-weights models, real-time GPU serving, and high-throughput batch inference, collaborating with teams across DoorDash, Wolt, and Deliveroo to ship reliable, scalable AI-powered features.

The role emphasizes cost-perf improvements, observability, and operator-friendly platforms, enabling rapid experimentation while meeting production standards for latency and

Qualifications

• Strong backend engineering fundamentals, especially in Python and distributed systems.
• Hands-on experience with LLM inference and/or fine-tuning of open-weight models in production.

Responsibilities

• Build production infrastructure for Generative AI including real-time GPU serving and high-throughput batch inference.
• Work across model serving, inference engines, training pipelines, and observability.
• Design scalable systems for model serving, batch inference, GPU autoscaling, and fine-tuning.
• Push cost and latency improvements for GPU inference while ensuring reliability and observability.
• Collaborate with ML engineers, product engineers, data scientists, and platform teams across DoorDash, Wolt, and Deliveroo.

Skills

Backend engineering

Python

Distributed systems

LLM inference

Fine-tuning

GPU utilization

Observability

Education

BSc/MSc/PhD in Computer Science or equivalent

Tools

Kubernetes

AWS

GCP

LLM serving frameworks

Job description

Deliveroo is seeking a Software Engineer for the GenAI Platform to build production infrastructure for Generative AI. You will work on open-weights models, real-time GPU serving, and high-throughput batch inference, collaborating with teams across DoorDash, Wolt, and Deliveroo to ship reliable, scalable AI-powered features.

The role emphasizes cost-perf improvements, observability, and operator-friendly platforms, enabling rapid experimentation while meeting production standards for latency and