About the job
Intercom builds AI-powered customer service tools for businesses around the world. Our flagship AI agent, Fin, helps companies deliver 24/7 support and handle complex questions that sometimes require a human touch. Integrated with our Helpdesk, Fin forms part of the Intercom Customer Service Suite, which supports nearly 30,000 businesses globally.
Founded in 2011, Intercom continues to set new standards for customer service by moving quickly, pushing boundaries, and delivering real value to clients.
Role Overview
The AI Infrastructure team at Intercom is looking for Senior AI Infrastructure Engineers in London. This group develops and maintains the systems that train and serve the next generation of Intercom’s AI models. The team’s work ranges from optimizing GPU architecture to building user-facing agents that handle millions of support requests each month.
Engineers here design the training pipelines and manage inference for custom models like Fin Apex, which leads the industry in customer service performance. The team operates at the core of Intercom’s AI efforts.
What You Will Do
- Design and scale training pipelines for large transformer and LLM models, including data ingestion, preprocessing, distributed training, and evaluation.
- Develop and improve inference services to deliver low-latency, reliable user experiences, covering autoscaling, routing, and fallback strategies.
- Optimize GPU-level performance by tuning kernels, improving utilization, and identifying bottlenecks in training and inference systems.
- Work closely with ML scientists to deploy advanced training and inference techniques.
What We’re Looking For
- Hands-on experience with model training or model inference at scale, or low-level GPU programming (such as CUDA or Triton). Experience in more than one of these areas is especially valued.
