Role overview

Cohere is an enterprise AI company building foundation models and products for real-world business applications. As the founding lead of Agentic Analytics, you will establish a new team within the Analytics and Data Insights organization responsible for creating an agentic system that answers business questions from Cohere’s internal data with verified sources, query transparency, and automatic escalation when confidence is insufficient.

Key responsibilities

Own the complete system architecture that transforms business questions into governed queries and verified answers, including question interpretation, context retrieval, definition enforcement, permission enforcement, and evidence attachment.
Design and implement core production services including agent-facing data tools, context retrieval systems, agent skills, and APIs that serve answers to platforms where teams already work such as Slack and IDEs.
Build and maintain an evaluation harness that defines correctness for each domain, including ground-truth question sets, result-level scoring, coverage metrics, and regression testing that blocks shipment of substandard releases.
Engineer the context layer by converting metric definitions, approved data joins, documentation, business glossaries, and examples into usable agent context, with every correction becoming a reviewed definition and regression test.
Drive production rollout across Cohere by bringing marketing, revenue, finance, and product domains onto the system sequentially, each with named owners, certified definitions, and evaluation sets.
Create replicable processes and documentation that enable other Cohere teams to operate the system independently.
Hire, lead, and develop the founding team while setting technical standards and conducting code review.

Requirements

Ten or more years of experience building and operating production analytics or software systems, with technical leadership spanning teams or large-scale systems.
Demonstrated production experience shipping LLM agents including tool use, retrieval and context engineering, guardrails, and observability, with clear understanding of common failure modes.
Hands-on background building natural-language or agentic analytics systems such as text-to-SQL or question-answering systems over semantic layers that served real users, with evidence of correctness.
Deep expertise in LLM system evaluation including golden datasets, result-level comparison, abstention metrics, error taxonomies, and regression testing in CI, with judgment to distinguish demonstrations from production-ready systems.
Strong Python and SQL skills with experience designing and running production services and APIs on major cloud platforms, particularly Google Cloud.

What makes this role worth considering

This is a hands-on staff engineering position where you will write production code and evaluation sets while building a new foundational system from the ground up. You will work on a problem that remains genuinely unsolved—ensuring agentic analytics systems provide correct answers rather than plausible wrong ones—and establish practices that become the model for how other teams implement similar systems across Cohere.