Staff Software Engineer - AI Agents

3w3 weeks ago

Traba

New York City, US · Full-time · $240,000 – $300,000

About this role

Traba is the AI operating layer for the industrial supply chain, building applied AI that drives real productivity gains. We started in workforce—temp staffing—and used it to embed ourselves inside daily operations. We are seeking an entrepreneurial Staff Applied Agent Engineer to lead development of our agentic platform as a founding team member.

You'll architect the agent platform end-to-end, from orchestration runtime to eval and observability stack. You'll partner with the CTO on core architectural decisions and spend real time in the field with customers and operators. You'll set the bar for quality and reliability by writing canonical examples.

You'll own foundational technical decisions including model strategy, agent harness design, and retrieval architecture. You'll bring the outside perspective on agent-building and data-product craft that this 0-to-1 product needs. You will build evaluation as a real engineering discipline with datasets and regression suites.

As a founding member of the Agents team, you'll raise the engineering bar and set the direction by shipping. Your work will directly impact how the global supply chain operates at scale, transforming customer operations through applied AI. This role offers the opportunity to shape a platform from the ground up.

Requirements

You've shipped agent systems into production at meaningful scale—designed the harness, picked the orchestration patterns, owned the evals, and lived with the on-call.
You have strong opinions on where to draw the line between prompting, fine-tuning, retrieval, and code, and why.
You are as comfortable in a warehouse on a customer site as you are in a design doc; you learn industry operations like WMS quirks, shift cadence, exception handling.
Ideal: prior experience at a vertical AI or AI-agent company (Nash, HappyRobot, Augment, etc.), Palantir/Scale-style FDE, or an AI-native data company (Hex, Omni, dbt).
You set direction by shipping, raising the engineering bar by writing the canonical example, not just the doc.
You have strong opinions on design patterns, eval datasets, prompt versioning, observability for agents, and the difference between a clean abstraction and an over-engineered one.

Responsibilities

Architect Traba's agent platform end-to-end—the orchestration runtime, the eval and observability stack, the integration layer to internal services and customer systems (WMS/TMS/ERP), and the patterns that every agent we ship is built on.
Own the foundational technical decisions for this part of the business: model strategy, agent harness design, retrieval and memory architecture, tool/MCP surface, and how we measure quality.
Spend real time in the field with customers and operators. Translate what you see into durable product and codify repeatable deployment patterns so each rollout compounds on the last.
Build evaluation as a real engineering discipline—datasets, graders, regression suites, experimentation tooling—so we ship agent improvements with the same confidence we'd ship any other software.
Partner with the CTO to make core architectural calls on how agents are built, evaluated, and deployed at Traba.
Set the bar for quality and reliability for the agentic platform.