Agentic Platform Engineer

Agentic Platform Engineer

Build and operate the execution systems that power real-world logistics workflows — workflow orchestration, durable event-driven pipelines, and safe agentic behaviors across the Navion platform.

Navion helps logistics operators move from reactive, labor-heavy workflows toward more efficient, increasingly agentic operations. The Navion Logistics Platform removes friction across the order-to-cash workflow by connecting planning, ordering, execution, visibility, and billing into a more continuous operational flow.

Our platform includes four connected components:
Keep Fill (Inventory Intelligence & Automated Replenishment), SPOT (Unstructured Order Intake Turned into Actionable Work), Portal (Real-Time Visibility Across Orders and Documents), and Back Office (Order-to-Invoice Reconciliation and Billing Readiness).

Navion works alongside existing TMS and document systems, helping carriers reduce manual work, improve workflow continuity, and scale without adding unnecessary complexity.

As we continue building the operational infrastructure for increasingly agentic logistics workflows, we are creating the execution layer that allows AI-driven systems to operate in controlled, reliable, production-grade environments.

Navion is seeking an Agentic Platform Engineer to build and operate the execution systems that power real-world logistics workflows across the Navion platform.

This role sits at the intersection of systems engineering, workflow orchestration, and applied AI. The Agentic Platform Engineer will help design the durable operational layer that turns unstructured inputs into structured state, coordinates work across systems and people, and enables safe agentic execution inside dispatch, visibility, and back‑office workflows.

This is a foundational engineering role for Navion's next phase of platform development. It is ideal for someone who thinks in systems, designs for reliability, and wants to help define how AI can move real operational work forward under practical constraints.

This is the most AI‑native role on the team. The Agentic Platform Engineer we hire already builds with agents daily — uses Claude Code (or equivalent) as their primary development environment, has shipped production systems with tool use, structured outputs, and multi‑step agent loops, and holds real opinions on models, context engineering, evals, and where agents belong versus where deterministic code is still the right answer.

We are not hiring someone who thinks LLMs are interesting. We are hiring someone operating at the frontier of what agents can reliably do — and who wants to push that frontier further inside real business operations.

• Build workflow orchestration systems that coordinate work across Navion's platform modules and external systems.
• Design durable, event‑driven pipelines that support complex operational workflows at production scale.
• Transform unstructured inputs such as emails, PDFs, texts, and documents into structured, actionable system events.
• Implement agentic workflow behaviors including planning, validation, retries, escalation paths, and human‑in‑the‑loop controls.
• Define system contracts, operational invariants, and failure‑handling patterns that support reliability and trust.
• Ensure strong observability across workflows through logging, tracing, replayability, and operational diagnostics.
• Help unify Keep Fill, SPOT, Portal, and Back Office into a cohesive execution layer across the Navion Logistics Platform.
• Collaborate closely with engineering, product, deployment, and customer‑facing teams to ensure systems are built for real operational conditions.
• Contribute to the technical architecture that enables Navion customers to move from manual coordination toward increasingly agentic operations.
• Design and maintain the eval infrastructure that keeps agentic behaviors trustworthy in production — golden datasets, regression tests, LLM‑as‑judge patterns, and drift monitoring as models change underneath us.
• Make opinionated calls about where LLMs belong versus where typed schemas, deterministic code, or traditional ML are the right tool — and…