Job description

About Kovari Kovari develops frontier technology to transform physical businesses. We partner with established operators to change how they run, and we build and operate businesses of our own through our subsidiaries. We are a small team of builders, operators, and researchers. We deploy on site, alongside the businesses we serve, and iterate in the field. We have raised over $6M from leading investors. We are based in San Francisco and work in person. The Role You will research and ship manipulation policies that work reliably in the field. You own the loop from data to model to deployment, and you see your policies run on robots in live operations. What You'll Do - Research and develop high-reliability manipulation policies built for fast deployment and iteration - Operate in a fast data flywheel across multiple modalities - Deep debug failure modes in transformer and diffusion policies deployed in the field - Optimize policies for real-time (~10 Hz) inference on edge hardware What You Bring - Experience deploying robot policies on hardware (model-based, RL, or imitation learning) - Sim-to-real or real robot data experience - Experience with multimodal inputs: vision, depth, force/torque, proprioception - Edge inference optimization with TensorRT, ONNX Runtime, or TVM - CUDA kernel optimization - Bonus: publications at CoRL, RSS, ICRA, NeurIPS, or similar - Bias toward action: train fast, deploy early, learn from failure Values - Pace of learning trumps everything. - Relentless refinement of our craft. - Low ego, high ownership. - All-in on the mission. We work in person, and this isn't a 9-to-5.

Requirements

Must-haves

You will research and ship manipulation policies that work reliably in the field. You own the loop from data to model to deployment, and you see your policies run on robots in live operations.
Research and develop high-reliability manipulation policies built for fast deployment and iteration
Operate in a fast data flywheel across multiple modalities
Deep debug failure modes in transformer and diffusion policies deployed in the field
Optimize policies for real-time (~10 Hz) inference on edge hardware
Experience deploying robot policies on hardware (model-based, RL, or imitation learning)
Sim-to-real or real robot data experience
Experience with multimodal inputs: vision, depth, force/torque, proprioception
Edge inference optimization with TensorRT, ONNX Runtime, or TVM
CUDA kernel optimization
Bias toward action: train fast, deploy early, learn from failure
Pace of learning trumps everything.
Relentless refinement of our craft.
Low ego, high ownership.
All-in on the mission. We work in person, and this isn't a 9-to-5.

Nice-to-haves

Bonus: publications at CoRL, RSS, ICRA, NeurIPS, or similar

Application prep

Use the role’s language and application details while tailoring your materials.

Platform: ashby
Effort: quick

Benefits

Benefits detected in the listing.

Vision