Agentic Platform & Intelligent Systems

Role Overview

Zinnia is building a shared AI foundation that embeds agentic intelligence directly into enterprise workflows, transaction systems, and customer interactions. We are not building isolated AI features. We are building a reusable, governed platform that powers automation, decisioning, and intelligent assistance across the organization.

We are seeking a Senior AI Engineer who operates at the intersection of research and production engineering. You will design, rigorously evaluate, and productionize agentic systems that are measurable, reliable, and safe to deploy in a regulated environment.

WHAT YOU’LL DO:

• You will help design and implement the core architecture of our AI agentic platform. This includes orchestration frameworks for multi-step, tool-using agents; retrieval systems that unify structured and unstructured enterprise knowledge; and infrastructure that makes model behavior testable, reproducible, and observable.
• You will contribute to agentic transaction processing systems that embed AI directly into operational workflows — enabling classification, validation, routing, and automated task completion. You will also support the development of a unified intelligent agent network that serves multiple user experience personas from a single-governed foundation.
• You will build the experimentation backbone that ensures every AI capability is measurable. This includes designing offline evaluation pipelines, maintaining regression test suites for non-deterministic systems, and implementing backtesting frameworks to compare models, embeddings, prompts, and orchestration strategies.
• You will design and execute controlled A/B tests in production and define statistical guardrails for AI/ML model promotion. Improvements must be demonstrated through measurable lift — not anecdotal wins.
• You will implement continuous monitoring systems that track accuracy, confidence, grounding fidelity, latency, cost, and drift. Regressions must be detected early. System behavior must be auditable.
• You will help establish reusable components and standards that enable teams to build on the platform without duplicating logic or fragmenting architecture.

WHAT YOU’LL NEED:

• You have at least five years of experience building production software systems and meaningful experience deploying LLM-based or agentic systems in real-world environments.
• You have at least 2 years of experience implementing Retrieval-Augmented Generation (RAG) systems and understand the tradeoffs in chunking, embedding strategies, hybrid retrieval, re-ranking, and grounding evaluation.
• You have hands-on background with MCP (Model Context Protocol) Architecture/Servers, knowledge Graphs
• You have 1 year of experience building or significantly contributing to multi-step agentic workflows involving tool execution, planning, orchestration, or transactional automation.
• You have at least 2 years of experience designing evaluation frameworks for AI systems and are comfortable with statistical testing, experiment design, and interpreting noisy performance signals. You understand the limitations of automated grading and the risks of benchmark overfitting.
• You have experience running A/B experiments in production systems and defining decision thresholds grounded in measurable impact.
• You are highly proficient in Python and comfortable building cloud-native distributed systems with strong observability and versioning practices. Python (FastAPI, Pydantic, async) or TypeScript/Node (Express/Fastify/Next API routes); testing (pytest/jest), Git/PR hygiene, CI/CD.
• Implement LLM evaluation & guardrails: prompt/unit evals, Ragas, Langfuse, LangSmith, A/B tests, hallucination & safety checks, feedback loops.
• You understand the governance and risk implications of deploying AI systems in regulated environments and can design for auditability and control from day one.

What Success Looks Like:

Agentic components are reused across workflows rather than rebuilt for each use case. AI-driven automation measurably increases straight-through processing and reduces manual intervention. Model and agent updates are evaluated against shared benchmarks before release. A/B experiments demonstrate statistically significant improvements prior to scale. Regressions are detected automatically. Performance, cost, and risk are continuously monitored.

WHAT’S IN IT FOR YOU?

At Zinnia, you collaborate with smart, creative professionals who are dedicated to delivering cutting-edge technologies, deeper data insights, and enhanced services to transform how insurance is done. Visit our website at www.zinnia.com for more information. Apply by completing the online application on the careers section of our website. We are an Equal Opportunity employer committed to a diverse workforce. We do not discriminate based on race, religion, colour, national origin, gender, sexual orientation, age, marital status, veteran status, or disability.

 

_#LI-RS1