Founding Engineer, Robot Behavior and Decision-Making

The Bigger Picture

At Andromeda, we're building the social intelligence layer for humanoid robots - technology that allows robots to understand people, build relationships and become a meaningful part of everyday life.

Our first robot, Abi, is already doing this in aged care and healthcare, where she brings conversation, companionship and engagement to people every day.

Now we're scaling from early deployments into a platform that can operate across thousands of real-world environments. We're looking for individuals who want to tackle hard problems across robotics and AI, and help define what it means for robots to live and interact alongside people.

We are looking for an engineer to found our Mission Autonomy team, which will own our robot’s long-duration decision-making layer and orchestrate the totality of Abi’s day. Mission Autonomy continuously decides what Abi should do next, balances task-level trade-offs, and adapts the robot’s plan across an eight-hour shift. It sits above and orchestrates Abi’s existing layers that perform conversations, waypoint navigation and obstacle avoidance, perception, gestures, and operator tooling.

As the founding Mission Autonomy team member, you will propose and build Andromeda’s Mission Autonomy (MA) architecture from the ground up, using frontier models and evaluation (both sim and real-world) to train and launch MA, and work with our voice agent, mobile autonomy, and gestures teams to design harnesses through which MA can execute behaviors on Abi. Once built, MA should guide Abi through a home, choose when to join an interaction, keep her character, and recover when the plan breaks.

There is no playbook for this. You’ll need a keen eye for model evaluation and trade‑offs, balancing architectural leadership with strategic hands‑on work. The work combines research with practical robotics.

• Design the system. Challenge our initial concepts as you discover what a full shift actually needs: is it formulated as an agent harness ingesting scene graphs and acting over robust, typed action primitives? Or will you invoke World Models and VLA components? You will shape the Mission Autonomy roadmap for Andromeda.
• Prototype the system. Get a mission autonomy loop running in simulation early. Use it to decide what we build, before the company commits teams to an architecture.
• Co‑design and build simulation capabilities with our simulation engineer.
• Own the technical direction and run the design conversations with autonomy, chatbot, perception, audio and platform teams. Turn them into contracts those teams can build against.
• Shape the action primitives. Work out what the mission needs from the layers underneath. Specify it properly. Get it built, whether that means convincing the owning team or writing it yourself.
• Design for degradation. Define how Abi should behave when audio, perception, connectivity, navigation, or platform health degrades. Make those moments safe, useful, and legible to operators, while preserving Abi’s character rather than treating every failure as a generic error state.
• Establish deployment patterns. Create a repeatable path from simulation prototypes to demonstrations on the robot, with the configuration, observability, test gates, and rollback mechanisms needed to iterate safely.
• Build the evaluation system. Define benchmarking protocols, evaluation methods, and reporting that let us continuously measure the longest‑horizon missions Abi can complete, the interventions they require, and the failure modes that limit them.

Month 1: Learn about Abi in the real world: read the Concept of Operations document, challenge its assumptions, and spend time with Abi in an aged‑care home. Bring the autonomy, conversational AI, perception, audio, platform, and simulation teams together to define how the system makes decisions and how the layers work together. Deliver the first architecture and a clear six‑month build plan, including what not to build.

Month 2: Build a working slice in simulation:
Abi drives to a resident’s room, arrives safely, and starts a conversation. Use simulation to discover what breaks when the…