Senior Data Scientist - AdTech

Senior Data Scientist - Ad Technology - Machine Learning

• Engagement Full-time, permanen
• tLevel Senior individual contributor / technical lead (no direct reports
• )Location New York City; onsite 2-3 days per wee
• kStart Immediate!!!! 🏃 ♂

• ️
• About the Clie
• ntOur client is operating at massive scale, the company processes enormous volumes of real-time bidding data and makes millisecond ad-serving decisions across mobile and CTV inventory. Machine learning is its core competitive advantage: proprietary models directly optimize advertiser outcomes, and the company is expanding its ML organization in New Yor

• k.
• Role Summ
• aryThis is a science-first, analysis-driven role for someone who lives inside the model rather than around it. You will own the investigative work that keeps the company's core prediction models performing: finding where models underperform, forming and testing hypotheses about why, and translating what you learn into concrete, prioritized recommendations that the ML engineering team implements. This is deliberately not a model-building or infrastructure role. You will not design new architectures or refactor production systems. You will diagnose, experiment, and explain, and your insights will set the roadmap for a multi-billion-dollar optimization probl

• em.
• What You'l
• l DoDiagnose and Improve Core Mo
• delsInvestigate where the install-probability prediction models lose accuracy, drilling into specific data slices, segments, and campaign types to isolate where and why performance degra
• des.Run calibration checks, feature-importance analysis, and slice-level error analysis to turn a vague performance gap into a precise, testable explanat
• ion.Form hypotheses about model behavior and validate them empirically: for example, running an existing pipeline on a reduced sample to test a data-volume hypothesis, then reasoning about what the result implies interna

• lly.
• Turn Analysis into A
• ctionTranslate findings into clear, prioritized recommendations and hand them to the ML engineering team to implement, defining the experiment and the expected mechanism, not just the me
• tric.Own reactive investigations end to end: when a campaign's performance drops in production, independently determine the root cause (data, model shift, prediction drift, privacy-driven signal loss) and report back with a diagn
• osis.Surface longer-term opportunities across the model portfolio, quantifying the potential lift and the trade-offs invo

• lved.
• Communicate and Lead Through Inf
• luenceProvide proactive status updates and generated hypotheses to technical leadership without being pro
• mpted.Operate with a high degree of autonomy: take a loosely defined problem and drive it to an answer independ

• ently.
• Required Qualifi
• cationsMaster's degree in data science, statistics, applied mathematics, computer science, or a related quantitative field (PhD pref
• erred).Senior-level experience applying data science or applied ML to large-scale prediction problems; strong candidates with roughly three-plus years of exceptional industry experience will be cons
• idered.Deep, low-level reasoning about model internals. You can explain not just that a change improved a metric, but why, in terms of what it does to the model's weights, loss surface, and predi
• ctions.Strong applied statistics and probability, including experiment design and the ability to reason about calibration, class imbalance, and evaluation on skewe
• d data.Working proficiency in Python and SQL, enough to run and modify existing pipelines and experiments (you will touch code to test hypotheses, though you will not own production sy
• stems).Autonomy and initiative. You can be handed a problem and trusted to drive it without day-to-day ove

• rsight.
• Ideal Candidate
• ProfileIntensely curious about how models actually work; you instinctively slice the data and ask "why" before reaching for a new te
• chnique.Scientific rather than mechanical: you treat a high accuracy number as a question, not an
• answer.Comfortable working at a low level of detail; you are wary of staying at the architecture-diagram altitude when the answer lives in t
• he data.Prior experience in advertising technology or another large-scale, real-time prediction domain is a strong plus, though raw analytical ability and scientific mindset matter more than domain bac
• kground.Self-directed and communicative: you keep leadership informed and raise hypotheses and risk

s early.