Robot Learning Engineer

Haptica Robotics • San Francisco, CA

3d • Onsite

About The Position

Haptica is seeking a Robot Learning Engineer to lead the entire training function. While Haptica does not deploy robots, the company requires in-house training capabilities to assess its effectiveness. This role involves creating the training pipeline, rigorously testing sensors, determining optimal data collection strategies, evaluating policies on robots, and assisting customers with data integration. The position offers significant autonomy and builds upon existing foundational work, including a robot arm, basic glove-based hand-tracking, and preliminary policy training groundwork. The initial focus will be on building the internal data collection pipeline with the hardware team, driving research on the impact of tactile data in learning, and shaping the roadmap for customer data utilization.

Requirements

• MS or PhD in robotics, CS, EE or a related field, or equivalent industry experience.
• Extensive experience in production-level software and ML engineering best practices.
• Experience with modern deep learning frameworks (e.g., PyTorch).
• Outstanding written and verbal communication skills, skilled at simplifying complex topics.
• Knowledge of the fundamentals for running imitation learning (diffusion policy, ACT or similar) from data collection to training, deployment, and debugging.
• Comfortable with hardware, including experience with cameras, IMUs, calibration, and time synchronization across sensors.
• Comfortable with ambiguity, building from scratch, defining constraints, and closely partnering with internal and external teams.

Nice To Haves

• Experience with dexterous manipulation, multi-fingered hands, or retargeting
• Passion for tactile sensing with relevant research or experience
• Experience taking egocentric or wearable data through to a trained policy
• Strong publication record in a relevant field
• Prior work on humanoids or highly dexterous robotic platforms

Responsibilities

• Build our data collection, training and inference pipeline on off-the-shelf hardware integrated with Haptica.
• Train imitation learning policies (ACT, diffusion policy) with and without our tactile data on tasks where touch matters most: shear, occluded grasps, delicate objects.
• Define how we evaluate. Benchmark tasks that reuse publicly researched setups, metrics, and tooling to quantify the impact of our data and the variance of the impact across tasks.
• Extend data collection from teleop to egocentric glove data. Integrate hand tracking, test retargeting from a human hand to grippers and multi fingered hands, and run experiments that decide how many fingers we track.
• Influence hardware decisions based on what our sensing needs to deliver for learning. You'll have opinions on sampling rate, resolution, synchronization, drift, and calibration.
• Stay on the bleeding edge of how the world's best companies are running robot learning, and help us integrate Haptica into customers' data collection and teleoperation setup.
• Share your findings with the world through demo videos, writeups, and papers with our academic collaborators at Stanford, MIT, and more.

What This Job Offers

• Job Type: Full-time
• Career Level: Senior