ABOUT THE ROLE

We are looking for a passionate and technically strong Robotics Manipulation Engineer to join our team at VinRobotics. You will work on the full stack of robot manipulation — from high-level task planning with behavior trees down to low-level trajectory optimization, collision-aware planning, and learning-based policies. You will play a key role in designing, implementing, and validating manipulation capabilities on real robotic hardware.

This role is open to both junior and senior candidates. We care more about your depth of understanding and curiosity than years of experience.

RESPONSIBILITIES

(1) Motion Planning & Control

- Design and implement motion planning pipelines, including custom planners and post-processing stages.

- Tune and benchmark sampling-based and optimization-based planning algorithms for speed and reliability on real hardware.

- Maintain accurate collision models for the robot and its environment to enable safe, reliable execution.

(2) Task & Behavior Architecture

- Develop modular, reactive robot behaviors using behavior trees, including custom action and condition nodes.

- Design and maintain behavior architectures for complex, multi-step manipulation tasks.

(3) Learning-Based Manipulation

- Develop and integrate learning-based policies for manipulation tasks such as grasping, in-hand manipulation, and contact-rich assembly.

- Design and maintain data collection pipelines for robot learning, including demonstration collection and simulation-based data generation.

- Evaluate and benchmark learned policies in simulation and on real hardware, and develop strategies for sim-to-real transfer.

- Collaborate with the broader team to identify where learned behaviors complement or replace classical planning approaches.

(4) System Integration & Validation

- Integrate perception, grasping, planning, and control subsystems into end-to-end manipulation pipelines.

- Design and run simulation-to-real validation workflows to de-risk hardware deployment.

- Perform root cause analysis on hardware failures, planning failures, and runtime errors.

REQUIREMENTS

Junior Engineer (1–3 years experience)

- Solid understanding of ROS2 core concepts (nodes, topics, services, actions, TF2).

- Hands-on experience with MoveIt2 for basic motion planning tasks (joint-space and Cartesian planning).

- Familiarity with OMPL and awareness of common sampling-based planning algorithms (RRT, RRT*, BFMT*, CHOMP, STOMP).

- Working knowledge of C++ (C++17) and Python for robotics development.

- Exposure to BehaviorTree.cpp or similar behavior tree frameworks for task logic.

- Understanding of rigid body transforms and coordinate frames in 3D space.

- Familiarity with deep learning fundamentals and at least one major framework (PyTorch or TensorFlow).

- Awareness of imitation learning or reinforcement learning concepts, even if only through coursework or personal projects.

- Comfortable using Git in a team development environment.

Senior Engineer (4+ years experience)

- Deep expertise with MoveIt2 internals — custom planners, planning pipeline plugins, servo control.

- Strong understanding of OMPL planner configuration, benchmarking, and adaptation for constrained planning.

- Experience designing and tuning FCL collision geometry for complex robot/environment setups.

- Production-grade experience with BehaviorTree.cpp, including custom node types and tree lifecycle management.

- Hands-on experience training and deploying learning-based manipulation policies (e.g. imitation learning, reinforcement learning, diffusion policies) on real hardware.

- Experience building robot learning data pipelines, including teleoperation-based demonstration collection or large-scale simulation data generation.

- Familiarity with sim-to-real transfer techniques and the practical challenges of deploying learned policies outside simulation.

- Track record of delivering full manipulation pipelines from planning through hardware validation.

- Strong systems thinking: performance profiling, latency analysis, and failure mode handling.

- Ability to mentor junior engineers and contribute to architectural decisions.

Nice to Have

- Knowledge of grasp planning approaches such as analytic grasp synthesis or learned grasp generation.

- Knowledge of force/torque sensing and compliant or impedance control strategies.

- Knowledge of deformable object manipulation or cloth/flexible object handling.

- Experience with real-time systems and hardware communication protocols (e.g. EtherCAT).

- Contributions to open-source robotics projects (ROS2, MoveIt2, OMPL, BehaviorTree.cpp, etc.).

- Background in optimal control, trajectory optimization, or model-based reinforcement learning.

- Experience with simulation environments for robot learning (e.g. Isaac Sim, MuJoCo, Gazebo).