Robotics Software Engineer

Mutable Tactics is a UK-based, VC-backed deep defense-tech startup focused on the orchestration layer for autonomous defense robotics. Its flagship product, Mastermind is an edge orchestration layer that lets one operator command a heterogeneous team of third-party drones and ground/maritime robots, including in communications-degraded environments. Mastermind fuses sensor data from the whole team into a shared world model, plans at the symbolic level, and executes through behaviour trees on each platform. Pre-seed funded, small technical team, hardware in the field.

2. The role

You will own core Mastermind modules end to end and turn our decision engine — symbolic task planning, behaviour trees, probabilistic state estimation — from prototypes into deterministic, tested, real-time C++ that runs on edge compute and flies in trials.

2.1 What you will do

• Design, implement and own core modules in modern C++: task execution, world-state management, platform adapters.
• Work with the CTO on the decision engine: PDDL-based task planning, behaviour trees, planning under uncertainty. Take components from prototype to production.
• Build the reasoning side of the world model: consume fused tracks from our perception partner, carry their uncertainty into planning, and abstract them into the symbolic state the planner acts on.
• Define and maintain the interface contract with the perception partner: message schemas, timestamping, uncertainty representation, degraded-mode behaviour.
• Set the engineering bar with the chief architect: code review, test strategy, SITL/HIL in CI, profiling on target hardware.
• Integrate partner platforms (MAVLink, ROS 2, Cursor-on-Target) and run field trials. Expect regular UK travel and occasional international trips.
• Support junior engineers through review and pairing.

3. Must have

• Production experience in modern C++ (17/20): concurrency, memory ownership, performance profiling, unit/integration testing, CMake and cross-compilation. The interview includes a C++ exercise.
• Software you wrote has run on real robots or vehicles, not only in simulation.
• Hands-on experience with at least one of: symbolic/task planning, behaviour trees, MDP/POMDP decision making, reinforcement learning — and a view on when each is the wrong tool.
• Linux and ROS 2 or comparable middleware (DDS, Zenoh).
• Eligible for UK SC clearance.

4. Strong advantage

• Planning and decision making under uncertainty: belief states, POMDPs, probabilistic inference.
• Working knowledge of sensor fusion and multi-target tracking outputs (track states, covariances, track lifecycle) — enough to consume a fused picture and spot when it is wrong.
• ArduPilot or PX4, MAVLink, and outdoor flight testing.
• Cursor-on-Target / ATAK, STANAG 4586.
• Inference on edge hardware (Jetson, TensorRT).
• Simulation: Isaac Sim, Gazebo, Unity.

An MSc or PhD in robotics, AI or a related field is useful; equivalent industry experience counts the same.

5. What we offer

• Salary: very competitive market rates, plus meaningful early-stage equity.
• Location: Cambridge, hybrid; on-site 3 days a week.
• Direct ownership of core system components from day one.
• Holiday, pension, equipment budget, learning environment, top-tier team, perks, etc.