AI Infrastructure Engineer

The role

This role sits within the AI Infrastructure team at Fin, a Salesforce company building an AI customer agent. The team operates a small group that trains and serves Fin's proprietary models, working from GPU-level tuning up to the user-facing agent. The position is hybrid, with an expectation of regular in-office attendance.

What you would do

• Build end-to-end training pipelines for large language and transformer models, covering data prep, distributed runs, and evaluation
• Develop inference services emphasizing latency and uptime, including scaling, request routing, and failover paths
• Tune GPU kernels and raise hardware utilization by finding performance chokepoints across training and serving
• Partner with ML researchers to productionize new training and serving techniques
• Mentor engineers and take part in interviewing and hiring for the team
• Set and enforce standards for reliability and operational quality across the AI platform

What they asked for

• Five or more years of professional software engineering with a record of shipping robust products or platforms
• Hands-on work in model training or large-scale inference, ideally for transformers or LLMs, or low-level GPU programming
• Experience operating systems under real production load in terms of traffic, data volume, or team size
• Strong grasp of computer science fundamentals, typically from a relevant degree or equivalent practical depth
• Clear communicator who can explain technical material to engineers and non-engineers alike
• Solid command of at least one general-purpose programming language

Nice to have

• Background at AI-focused organizations that train or serve their own models
• Running training or inference jobs on Kubernetes
• Experience with AWS or another major cloud provider
• Production Python use in ML or infrastructure work
• Public evidence of technical interest such as side projects, open source, talks, or writing