Founding Robotics Engineer, Perception & VLA

Destro AI•New York, NY

About The Position

Destro AI is hiring a Founding Robotics Engineer to lead perception and robot learning for VisionOS, its robot-level intelligence product. This founding role involves building the perception and learning stack from the ground up, setting its technical direction, and working independently with the founding team. The role covers the full model lifecycle, from perception pipelines and vision-language-action (VLA) models through pre-training, fine-tuning, deployment, post-training optimization, and inference. Models are considered complete when they work reliably on customer floors, not merely when they perform well in evaluation. The engineer will develop systems for 3D scene understanding, product and case pose estimation, and zero-shot identification; build and train VLA models; deploy models on physical robots within latency and compute constraints; and create data collection and active learning workflows. The first month includes learning the robots, simulation environment, and customer floors, drafting a perception and learning plan, and getting an initial model running on a physical robot performing a real task. This is a full-time, travel-heavy role based in the New York/New Jersey area, with work on customer floors and significant time spent at customer warehouses. Compensation includes base pay and equity.

Requirements

3 to 5 years of hands-on work in robotics, computer vision, or machine learning, ideally on physical robots.
Hands-on model training on GPUs.
Experience shipping models to production inference under latency constraints.
3D perception experience with point clouds and depth cameras.
Experience deploying on real robot hardware and working in simulation.
Strong Python and C++ skills, or proficiency in a similar systems language.
Ability and inclination to get physical prototypes working in weeks rather than months.
Willingness to travel frequently and spend significant time on customer sites.

Nice To Haves

VLA or manipulation policy training.
Zero-shot pose estimation or segmentation.
Grasp planning, learned or geometric.
Warehouse or logistics automation experience.
Experience with MuJoCo, Isaac Sim, or similar simulation environments.

Responsibilities

Own perception and robot learning for VisionOS and set the technical direction for the perception and learning stack.
Build perception pipelines for 3D scene understanding, including RGB-D fusion, point clouds, segmentation, 6-DOF pose estimation, and zero-shot identification with vision-language models.
Develop VLA models, including pre-training, fine-tuning, and the use of synthetic data in simulation.
Deploy models on real robots within tight latency and compute constraints.
Build an active learning loop that turns mistakes observed on customer floors into improved models.
Set up teleoperation, collect data, and build a pipeline to turn raw recordings into clean, versioned training data.
Take models from simulation to physical robots and test and improve them in live customer operations.
Develop a perception and learning plan and get an initial model running on a physical robot during the first month.