Senior Staff Machine Learning Engineer

Grubhub•New York, NY

4d•Hybrid

About The Position

Grubhub is looking for a Senior Staff Machine Learning Engineer to help lead the machine learning engine behind Discovery: the ranking, recommendation, and retrieval systems that decide what every diner sees when they open the app or run a search. These models sit on the critical path to conversion for hundreds of thousands of merchants and hundreds of millions of menu items, and they are one of the largest organic growth levers we have. This is a hands-on technical leadership role, not a management role. You will own model architecture across several connected technical areas: search ranking, homepage and topic recommendations, retrieval, and query understanding. You will be accountable for how those pieces fit together, not only for any single model. You will set technical direction alongside other staff engineers, product managers, and platform partners. You will raise the bar on how the organization builds, evaluates, and operates models, and mentor the engineers around you through design reviews, code reviews, and direct feedback. We expect you to challenge technical decisions across teams when the engineering case is clear, and to bring evidence when you do. Our team practices end to end project ownership, and our work focuses heavily on personalized recommendation, retrieval, and classification from catalog content and clickstream. Deep neural networks, learned embeddings and approximate nearest neighbor retrieval, transfer learning from pre-trained large scale models, calibration, classic regressions, fine tuning, and large language models all have a place in our daily lexicon.

Requirements

MS/PhD in a quantitative discipline (Computer Science, Math, Physics, Engineering, Statistics or other technical field) or equivalent experience
8+ years building and shipping machine learning systems, including 3+ years operating at staff-level scope: setting technical direction across multiple teams, model families, or systems
Deep experience in recommendation systems, ranking, or information retrieval at scale, in production and under real latency and cost constraints
Proven track record with production deep learning in TensorFlow or PyTorch, including training, serving, and tuning runtime models on GPUs
Experience with Large Language Models and transformer-based architectures, including fine-tuning, embedding generation, and deploying them in latency-sensitive applications.
Experience with language understanding over imperfect grammar (real-world search queries, menu and catalog text) is a strong plus
Strong data engineering fundamentals: PySpark, Hive/SQL, the Python data stack, and feature pipelines you can debug as well as build
Fluency with experimentation: designing A/B tests, choosing the right guardrails, and recognizing when an offline metric is misleading you
Experience with cloud ML infrastructure (AWS/SageMaker or equivalent), model deployment, and production monitoring and observability
Demonstrated technical leadership: mentoring engineers, driving design and architecture reviews beyond your own team, and influencing decisions without direct authority
Comfort communicating performance metrics, model behavior, and technical trade-offs to both deeply technical and non-technical audiences, up to and including executive stakeholders
Ability to keep up with the latest research publications and synthesize them into working production systems
Deep interest in self-motivated continuous learning

Responsibilities

Own the architecture of our ranking and recommendation stack end to end: candidate retrieval, multi-objective ranking, calibration, and the ensemble that trades conversion against profitability.
Make the cross-system design calls that no individual model owner can make alone.
Lead the evolution of our optimization objective from short-term conversion toward long-term diner value, including the offline evaluation and online experimentation work required to trust the result before it ships.
Bring state of the art research in information retrieval and recommender systems into our runtime environment: LLM-driven query and intent understanding, embedding-based retrieval, sequential user representations for cold start, and real-time inference.
Assess rigorously what actually transfers to our traffic, and say no to what does not.
Raise engineering and operational standards across multiple teams: model evaluation and scorecards, reproducible training pipelines, safe deployment, SLOs and observability for tier-1 models, and proactive management of technical debt before it becomes urgent.
Partner with Product, Search Engineering, Ads, and Data Platform to shape roadmaps, surface risk early, and make sure the data and infrastructure exist before the model needs them.
Mentor senior and mid-level engineers, participate in hiring, and grow the technical depth of the team so that no critical system depends on a single person.
Translate technical trade-offs into business terms for product and executive stakeholders, and document the rationale clearly enough that decisions outlive the people who made them.
Question existing assumptions, look for the innovation we are not yet pursuing, and relentlessly analyze and improve the performance of our business.