Data Project Manager, Autopilot AI

Tesla

Tesla Autopilot applies deep learning for autonomous driving at scale. Past features we’ve launched include Navigate on Autopilot (making lane changes and taking highway exits), Smart Summon (calling the Tesla to you in a parking lot without anyone in the vehicle), Stopping for Stop Signs and Traffic Lights, Radar Removal, and more than a dozen active safety features. Currently, we are working to release the Full-Self-Driving feature on city streets. To achieve this, the team actively collects massive datasets from the customer fleet, creates ground truth of labeled videos, and deploys deep neural networks to our fleet of more than two million Tesla cars.

As Data Project Manager, you will run the “data engine,” the process by which we improve neural networks via data annotation by leveraging Tesla’s fleet. You will take ownership of the datasets of specific vision models (ex. stop signs) and make them large, high quality, and diverse. You will identify areas where neural networks are challenged (ex. even blue stop signs!), work with engineers to formalize requirements and deliverables for the data annotation team, and close the loop by driving improvements in models via data. You will manage the variety of datasets (what should get annotated), the ontology (how data should be labeled), and close the loop on the project’s impact on improving Autopilot performance and reducing interventions (why data is labeled). The data engine is the most crucial aspect of “pre-neural-network” development and has unlocked every major Autopilot customer release.

To learn more, check out these talks by Andrej Karpathy our Director of AI: ScaledML 2020 https://www.youtube.com/watch?v=hx7BXih7zx8 and AI Day: h ttps://www.youtube.com/watch?v=j0z4FweCy4M

You will:

  • Manage projects through ideation, development, pilot, implementation, and completion
  • Collaborate with neural network engineers to triage scenarios where neural networks are challenged and should be improved via data Leverage the Tesla fleet to source interesting, diverse, and high volume scenarios to annotate
  • Design annotation ontology with engineers, considering the implications for training neural networks
  • Partner with the data annotation leadership team to set goals, forecast timelines, and ensure goals are met
  • Design labeling interfaces by creating and testing core features, flagging issues, and prioritizing feature requests
  • Communicate project progress, risks, and successes to stakeholders
  • Dive into the details including annotating data yourself and testing out new features in labeling interfaces
  • Close the loop between training and evaluation suites to measure the impact of new data on the neural network
  • Verify new data improves in-car Autopilot performance by reducing interventions and communicating if the intended effect is not happening Identify process inefficiencies and improve workflows with Autopilot Leadership, Program Management, and Engineering teams

We are looking for candidates with:

  • Two+ years of industry experience in product, program, or project management
  • BS or advanced degree in engineering or analytics heavy major
  • Understanding of machine learning basics (training vs. test set, overfitting, precision/recall, etc.)
  • Excellent work ethic, communication, and collaboration skills
  • Ability to adapt quickly to a very fast pace environment. You like simultaneous opportunities and challenges in a rapidly evolving company Passion for the future of transportation and autonomous technologies