Senior Software Engineer (Agents)

Traba•San Francisco, CA

7d•$200,000 - $240,000

About The Position

Traba is the AI operating layer for the industrial supply chain. We started in workforce—temp staffing, the biggest operational pain point for the manufacturing and logistics customers we serve—and used it to embed ourselves inside their daily operations and create a far better customer experience through technology. Now those same customers are pulling us beyond staffing into the broader operational workflows that run their facilities. That foundation gave us proprietary data from millions of shifts and deep enterprise relationships. But our edge is more than data: by connecting to the systems running across every facility and activating the workers already on our platform to execute against them, we are building applied AI that drives real productivity gains and transforms how the global supply chain operates at scale. We are backed by Founders Fund, Khosla Ventures, and General Catalyst. We are seeking an entrepreneurial Senior Applied Agent Engineer to join as a founding member of the Agents team and help us build the next layer of Traba's product—an agentic platform that synthesizes the data flowing through our marketplace and operates autonomously inside our customers' supply chain workflows. You'll embed deeply in our customers' operations, design and ship production agent harnesses on top of frontier models, and turn what you learn in the field into durable product. You will partner closely with our CTO, product, and our operators to take 0-to-1 ideas to scaled deployments.

Requirements

5+ years of professional software engineering experience, with at least 1+ year shipping LLM- or agent-based features into production.
Strong proficiency in Python and/or TypeScript/Node.js, and comfort designing APIs, distributed systems, and data models in PostgreSQL.
Hands-on experience with the modern agent stack: prompt engineering at production scale, evaluation frameworks (e.g., Langfuse, Braintrust, internal harnesses), agent orchestration patterns, and frontier model APIs (Anthropic, OpenAI, etc.).
A track record of building and shipping in fast, ambiguous environments—ideally at an early-stage vertical AI, AI-agent, FDE-style (Palantir, Scale), or data-product company (Hex, Omni, dbt).
Excellent written and verbal communication—you can run a customer workshop in the morning and write a clean design doc in the afternoon.
A genuine excitement about industrial operations and the chance to use applied AI to make them dramatically better.

Nice To Haves

Builder with an AI operator's instinct. You've shipped real product on top of LLMs—not just chat wrappers. You've designed agent harnesses, structured tools, written evals, and tuned prompts against production traces. You think in terms of capability, reliability, and unit economics, not just “does the model say the right thing.”
Domain-immersed. You enjoy spending time with the people who actually do the work. You learn an industry's vocabulary, edge cases, and operational tempo, and let that knowledge shape what you build. Bonus points if you've done this for a vertical AI or AI-agent company before (e.g., Nash, HappyRobot, Augment, Pallet, Harvey, Legora, ElevenLabs).
High-agency in ambiguity. You're comfortable being dropped into a fuzzy problem with a customer, a half-formed hypothesis, and a deadline. You scope, build, evaluate, and ship without waiting for someone to hand you the spec.
Sweat both ends of the stack. You can move between prompt iteration, eval design, backend services, data modeling, and the customer-facing UI in the same week. You care about the small details—evals that actually catch regressions, prompts that fail gracefully, traces that are easy to debug.

Responsibilities

Embed with our customers and operators to deeply understand how supply chains run today—then design and ship agents that take meaningful work off their plate.
Build production agent systems on top of frontier LLMs: tool use, sub-agents, retrieval, structured outputs, MCP servers, and the orchestration layer that ties them together.
Own evaluation as a first-class discipline. Build datasets from real traces, write rubrics and graders, run experiments, and ship improvements that you can prove move the needle.
Architect and build the data, services, and APIs the agent layer depends on—integrating with our internal systems and our customers' WMS, TMS, and ERP environments.
Codify repeatable deployment patterns so each new customer rollout is faster and more reliable than the last.
Be a key player on a scrappy founding team—contributing to product direction, hiring, and the engineering culture as we grow.