Agentic AI Engineer

Location: Bristol, UK (Hybrid for top talent)

Team: Applied Research - Robotics & Intelligent Systems

Employment Type: Full-time, Permanent

About the Role

Kinisi’s KR-1 bimanual robot is evolving from a collection of dexterous skills into a true autonomous agent that is able to perceive, reason, plan, and act across diverse manipulation tasks. As an Agentic AI Engineer, you will bridge low-level learned behaviors with high-level cognitive reasoning, building multi-agent software that sequences, coordinates, and adapts skills in real time. Your work will unlock the next generation of versatile practical robots that can take on everything from precision assembly, pick and place, to high-force, whole-body manipulation.

What You’ll Do

• Agent Architecture & Reasoning
  • Design hierarchical or graph-based agent frameworks (e.g., Semantic Kernel, Smolagents, CrewAI, AutoGen-style tool-use) that reason about goals, context, and available skills.
  • Integrate large language and visuo-motor policies for task planning, constraint satisfaction, and error recovery.
• Skill Orchestration
  • Compose and schedule bimanual manipulation skills - learned from teleop, human demos, and simulation - into robust task graphs.
  • Develop meta-controllers that refine skills on-the-fly using sensory feedback and task progress.
• End-to-End Deployment
  • Write production-quality Python and C++ to train, evaluate, and deploy policies on KR-1 hardware.
  • Lead design reviews and collaborate with controls, perception, and cloud-infra teams to turn research prototypes into field-ready code.
  • Debug and drive continual learning.
  • Instrument, profile, and debug agentic systems running on live robots.
  • Build data-collection pipelines that surface failure cases and automatically trigger skill-training loops.

Required Qualifications

• M.S. in Computer Science, Machine Learning, Robotics, or related field (or equivalent practical experience).
• Hands-on experience training and deploying learned policies for complex behaviours in robots or simulated characters.
• Fluency with modern ML toolchains (PyTorch/JAX, RL or behaviour-cloning libraries, agentic frameworks, MCP, distributed training workflows).
• Solid software-engineering fundamentals - tested, documented, maintainable Python & C++.
• Demonstrated ability to debug integrated hardware-software systems.

Preferred Qualifications

• Ph.D. in CS, ML, Robotics, or related discipline.
• Track record with large-scale behaviour-cloning techniques (e.g., diffusion policies, ACT, Gato, Pi0).
• Prior contributions to sizable, shared codebases and CI/CD pipelines.
• Direct experience working with robot arms, grippers, or mobile manipulation platforms.
• Familiarity with multi-agent planning frameworks, LLM-powered tool-use, or cognitive architectures (e.g., hierarchical task networks, behaviour trees augmented with generative AI).

What Sets You Apart

You’re equally excited about high-level reasoning and low-level control: comfortable letting a transformer plan the next ten minutes while you optimise the last ten milliseconds. You can articulate a vision for agentic robotics - and deliver the robust code, experiments, and data that turn that vision into reality. Join us and push the frontier where generative AI meets embodied intelligence.

What We Offer

• Competitive salary, equity, and performance bonus.
• Comprehensive health, dental, and vision coverage.
• Annual conference budget & dedicated research time.
• Flexible hours and hybrid/remote-friendly culture.
• State-of-the-art lab space with collaborative, cross-disciplinary teams.