Staff AI/ML Engineer

Design and ship production AI systems — multi-agent orchestration, routing, and specialized agents that take a request and carry it through to a reliable outcome.

Automate manual operational work across onboarding, support, exceptions, and document/data understanding — turning processes that take hours or days into seconds.

Build the models behind the decisions — forecasting, prediction, matching/allocation, optimization, and reliability scoring that ground the product in data instead of guesswork, exposed as services the agent layer can call.

Design the learning loop. Instrument decisions and their outcomes so models continuously improve, with the data and evaluation infrastructure to support it.

Own reliability and evaluation. Build the eval harnesses, tracing, observability, and guardrails for complex AI workflows where mistakes carry real operational and financial consequences — and prove a model or agent beats the status quo before it ships.

Make the build-vs-rules calls — know when a model genuinely wins, when an agent is the right tool, and when a simple rule is the smarter answer.

Raise the bar and help the team grow — push our prototyping-to-production pipeline forward and mentor engineers as the AI team scales.

Background in logistics, supply chain, transportation, marketplaces, mobility, or fulfillment.

Operations research / optimization, or reinforcement learning / bandits for sequential decision-making.

Multimodal / document understanding, computer-use, or browser automation.

Real-time / streaming systems, feature stores, and production MLOps at scale.

Patents or peer-reviewed publications, or experience as an early/founding engineer.

We're remote with a few in-person offsites throughout the year. We move quickly, default to writing, and ship early to real customers before rolling things out broadly. AI is part of how we work every day, not a separate initiative — we expect people to experiment with new tools, share what they learn, and continuously rethink how great products get built. We care more about speed, ownership, and iteration than process for the sake of process.