Staff Software Development Test Engineer | AI Evaluation

This is one of the more interesting roles on the board today. Tekion is scaling from a handful of AI agents to more than a hundred across service, sales, finance and analytics, and this role builds the evaluation platform that lets every ML team measure whether those agents are actually any good. You design evaluation datasets, build automated scoring pipelines, and define quality metrics for accuracy, consistency and safety of AI generated output. It is offered as a shared platform service rather than per team tooling, which is the right architecture and rare to see staffed properly. Stated band is five to eight years. Good fit if you are a strong SDET or ML engineer who wants to own AI quality as a discipline.

Stated band is 5 to 8 years in SDET, quality engineering, ML engineering or data science, with hands on experience building evaluation or measurement systems. Note that the board title says Staff while the description opens by calling it a Senior SDET role, so the level naming is inconsistent inside the posting itself. The band is the reliable number. The team is Tekion's AI Platform group. The stated premise of the role is that evaluation is the backbone of trustworthy AI, and the concrete driver is scale: Tekion is going from a small number of AI agents to over a hundred across Service, Sales, F&I and Analytics, and every one of those teams needs a way to know whether its agent is improving or regressing. Your job is to build that capability once, as a shared platform service, rather than letting each team improvise. Day to day you would work with ML engineers, data scientists, the AI Platform team and product management to design evaluation datasets, build automated scoring pipelines, and define quality metrics that quantify accuracy, consistency and safety of AI generated output. Location is Bangalore HQ. Fit guidance: this is a genuinely emerging speciality and the experience transfers well, because almost every company shipping LLM features is about to discover it has no way to measure them. It suits someone who is comfortable being the person who defines what good means, and who can hold that line with ML teams who would rather ship. If you want conventional test automation, this is not that job.