Senior Applied Scientist

What we do:

Applied scientists at TTD work closely with engineering throughout the lifecycle of the product, from ideation to production and monitoring. Our applied scientists are end-to-end owners. You will participate actively in all aspects of designing, researching, building, and delivering data-focused products for our clients and traders. 

With this role in AI Lab, you will focus on designing and developing agentic AI applications and systems, such as The Trade Desk's AI Assistant. You will own a technical area end-to-end, from evaluation through production, while also raising the bar for the scientists and engineers around you. 

What you'll do:

• Drive the evaluation platform for the AI Assistant, including benchmark generation, human and AI-judge scoring, and scenario-driven testing aligned with human feedback. 

• Drive prompt and harness optimization, including automatic tuning and research, and own the release gate that decides what ships to traders and buyers. 

• Build and maintain the guardrail and safety framework for input and output checks across all agents in the AI Assistant. 

• Set the evaluation, best practices and training standards that other teams adopt as agentic AI spreads across the business. 

• Mentor the scientists and engineers around you, and contribute to the applied science roadmap for AI Lab, identifying high-impact problems and coordinating with engineering, AI Infrastructure, and Omnichannel opportunities. 

Who you are: 

• You have a sustained track record of turning research into agent capabilities that ship, and you think beyond the immediate task to understand why it matters. 
• You are a driver. You can scope a six-week delivery cycle (our current shipping cadence, matching the wider product organization), keep it moving, and know when shipping fast would compromise the quality of what goes out.
• You have a strong sense of data and model intuition. At our scale, and in a fast-moving agentic space, off-the-shelf approaches often do not hold up, so you can work from first principles to build evaluation, guardrails, and optimization that fit our systems.
• You have a product-focused and start-up mindset and enjoy moving quickly on innovative product ideas, not just the technical problems underneath them.
• You mentor by doing. You guide and teach, bringing others along with you.
• You have a habit of guiding AI-assisted development. You know the difference between blindly accepting what a model produces, refusing to use it at all, and guiding it to the right answer, and you consistently do the last of these.

What You Bring to the Table:

• An abundance of intellectual curiosity, and enthusiasm to learn and teach new techniques as agentic AI keeps changing.
• Comfort working on an agile, distributed team spanning multiple time zones, and the ability to communicate clearly across technical and non-technical audiences.
• 5+ years of experience (or 3+ with a PhD) in a data-driven or applied ML role, including leading technical projects end-to-end. An advanced degree in a quantitative field is useful but not required. What you can contribute matters more than how you got here.
• Hands-on experience with Python and a deep learning framework such as PyTorch, including training or fine-tuning models through approaches such as supervised fine-tuning or preference optimization (e.g., DPO)
• Experience with Databricks, Spark, Delta Lake, and Unity Catalog, strong SQL, and experience tracking experiments and models with MLflow or a similar tool
• Experience serving models in production (for example vLLM or Ray Serve) and building or orchestrating agents with frameworks such as LangGraph, DSPy, or MCP-based tooling
• Comfort writing production-quality, reviewable code, and working with confidence and without ego: you have strongly held, defensible ideas, and you are quick to help a teammate when something breaks
• A commitment to valuing, seeking out, and fostering diversity of background and perspective on your team
• You have a track record of turning recent agent or LLM research into solutions for real, messy problems, not just benchmarks.
• You have experience building evaluation harnesses, benchmark datasets, or scenario-driven testing for LLM-based systems.
• You have experience with routing, prompt optimization, or memory systems for agents operating at scale.
• You have hands-on experience with guardrail or safety systems for AI-facing products, for example Llama Guard, NeMo Guardrails, or similar frameworks.
• You have experience with orchestration or workflow tooling such as Ray, Airflow, or Dagster for production data and training pipelines. 

The Trade Desk does not accept unsolicited resumes from search firm recruiters. Fees will not be paid in the event a candidate submitted by a recruiter without an agreement in place is hired; such resumes will be deemed the sole property of The Trade Desk. The Trade Desk is an equal opportunity employer. All aspects of employment will be based on merit, competence, performance, and business needs. We do not discriminate on the basis of race, color, religion, marital status, age, national origin, ancestry, physical or mental disability, medical condition, pregnancy, genetic information, gender, sexual orientation, gender identity or expression, veteran status, or any other status protected under federal, state, or local law.