Data Scientist I

The Trade Desk The Trade Desk · Media · Ventura · Data Science

The Trade Desk is seeking a Data Scientist to develop algorithms for Audience Targeting using deep learning at scale. This role involves end-to-end ownership from ideation to productionalization and monitoring, focusing on training and inference in always-on production systems.

What you'd actually do

  1. development of algorithms that help power our Audience Targeting
  2. train and develop inference at scale
  3. parsing of noisy labels and data, working with large output spaces (thousands of labels), and unreliable ground truth
  4. participate actively in all aspects of designing, researching, building, and delivering data-focused products for our clients
  5. leading others

Skills

Required

  • BS/MS with 4+ years or a PhD with 2+ years of experience working in a DS or ML role that involves bringing products from ideation to production
  • deep learning
  • communicate with diverse stakeholders, making architecture recommendations, ensuring effective execution, and measuring quality of outcomes
  • Proficient in at least two of: R, Python, Java, or Scala. Expert at one of those.
  • Strong SQL skills

Nice to have

  • TensorFlow/PyTorch preferred
  • Experience running heavy workloads on a distributed computing cluster (especially EMR or Databricks), leveraging technologies like Spark to work with large datasets preferred
  • Proficient in software engineering concepts like containerization and version control preferred

What the JD emphasized

  • sustained track record of making significant, self-directed end-to-end contributions to complex multi-layered impactful machine learning projects
  • deep expertise working with embedded models in always-on production systems
  • product-focused mindset
  • work from first principles to develop solutions and adapt them to a unique environment

Other signals

  • deep learning algorithms
  • train and develop inference at scale
  • end-to-end owners
  • productionalization and monitoring
  • embedded models in always-on production systems