Senior ML Engineer - LLM Agents & Evaluation

DoorDash, Inc.; New York (NY); Hybrid; USD 137,000 - 299,000; Full time

Benefits offered by this job

Equity grants

Comprehensive benefits package

Parental leave (16 weeks)

Medical/Dental/Vision

Paid time off

Job summary

DoorDash is seeking a Senior Machine Learning Engineer to join Agentic Foundations, focusing on vertical agent development within New Verticals ML. You will own continuous improvement of flagship agents, starting with the Ask Assistant, using an eval-driven loop to make them faster, cheaper, and more reliable for millions of customers.

You’ll collaborate with teams on agent memory, representations, and post-training of small language models, bringing these insights into production agents and

Qualifications

• 3+ years of ML experience with hands-on work building and shipping LLM-based agents (tool use, context management, prompting, guardrails) and evals/data.
• Experience with post-training or fine-tuning open-weight models (e.g., SFT, preference optimization, or RL).
• Strong NLP/ML foundation; proficiency in Python and frameworks like PyTorch or TensorFlow.

Responsibilities

• Raise the quality of Consumer agents by building eval harness and optimizing loop, improving prompts, tools, and context.
• Balance quality, latency, and cost with rigorous experiments and SLM deployment.
• Launch new agents for Merchants and Dashers from exploration to production.
• Turn foundations into product impact with memory and product representations in production agents.
• Shape the ML-driven roadmap for grocery and retail delivery growth.

Skills

LLM agents

Eval loops

Python

PyTorch

Open-weight models

Fine-tuning

Education

MS/PhD in CS or related

Tools

Post-training methods

SFT

RL

Job description

DoorDash is seeking a Senior Machine Learning Engineer to join Agentic Foundations, focusing on vertical agent development within New Verticals ML. You will own continuous improvement of flagship agents, starting with the Ask Assistant, using an eval-driven loop to make them faster, cheaper, and more reliable for millions of customers.

You’ll collaborate with teams on agent memory, representations, and post-training of small language models, bringing these insights into production agents and