Full Job Description
About the team
The AI/ML team within eBay's Selling organization builds machine learning and generative AI systems that help sellers make confident pricing and listing decisions. We combine rigorous econometric methodology with production ML engineering - moving research from theoretical grounding through reproducible prototypes to seller-facing features used across eBay's global marketplace.
About the role
As a Hands-On Senior Manager, Applied Research, you will design and advance our research agenda across machine learning and generative AI, contributing to our science work - while developing and growing the researchers who do that work with you.
This is a player-coach role. About 70% of your time is hands-on applied research: designing studies, building models and prototypes, writing production code, and taking your own research from prototype to deployed system with accountability for measurable business impact. About 30% is people leadership: managing a small but highly talented team of Applied Researchers , and - above all - developing them as researchers and as professionals.
You will partner closely with the technical architect who sets the technical direction for the team and is responsible for the technical decision making. You will also collaborate with product managers, data, backend and client side engineers to deliver customer facing deliverables.
You will work hands-on in building solutions for challenging problems around listing price and quality, listing recommendations, seller segmentation, agentic commerce to name a few.
You will also be responsible for fostering a team culture that encourages innovation and be able to coach your team members by gathering and providing feedback and helping them navigate career paths.
Qualifications
- Expertise in causal inference for ML: double machine learning, heterogeneous treatment effect estimation, observational study design, confounder handling.
- Hands-on experience with gradient boosting (LightGBM, XGBoost, CatBoost), including custom loss function design.
- Proficiency in advanced prompt engineering techniques: Chain-of-Thought, ReAct, few-shot calibration, structured output contracts, and context caching.
- Experience designing and deploying LLM-based classification or generation systems in production environments.
- Ability to build LLM evaluation frameworks: synthetic dataset construction, inter-annotator agreement measurement, and LLM-as-judge methodology.
- Demonstrated experience with agentic system patterns: tool-calling architectures, intent routing, multi-step reasoning, and safety/guardrail layers.
- Production-level experience taking ML research from prototype to deployed system, with accountability for KPIs and measurable business impact.
- Proficiency in Python with a modern data engineering stack (polars/pandas, pyarrow, scipy/statsmodels); familiarity with LangChain or an equivalent LLM orchestration framework.
- People management experience in managing applied researchers or data scientists.
- PhD in Computer Science, Machine Learning, Artificial Intelligence, Statistics, or a related quantitative field, plus 3+ years of industry research experience; or MS in a related field with 5+ years of applied ML research experience demonstrating equivalent depth.
Bonus Points
- Experience with survival analysis or competing-risks models in marketplace or platform settings.
- Familiarity with knowledge distillation: training lightweight encoder models (BERT-class or smaller) from LLM teacher signal.
- Hands-on work with synthetic data generation pipelines for NLP: stratified sampling, proportional quota design, and silver-to-gold adjudication.
- Experience with marketplace or e-commerce economics: price elasticity estimation, two-sided market dynamics, behavioral pricing.
- Track record of applying academic ML/NLP literature to production systems.
The character & qualities that will help you succeed:
- You're energized by the gap between research prototype and production system - you close it.
- You read papers and ship code; you know which problems need first principles and which need engineering judgment.
- You earn cross-team credibility by being right frequently and communicating it clearly.
- You mentor junior researchers not because it's required, but because their velocity is your velocity.
Additional Details
This job posting relates to an existing vacancy within eBay.