Director of Data Science, Commerce Site Lead

About the role

InMobi's Demand-Side Platform (DSP) runs real-time bidding at millions of queries per second under sub-50ms latency, and we are in the middle of a commerce-first re-architecture: moving the platform away from impression-and-click optimization toward driving measurable downstream commerce outcomes — purchases and return on ad spend (ROAS) — for advertisers.

The trajectory is the one the best-in-class performance-commerce platforms have already proven: a single AI engine that optimizes bidding directly to conversion and ROAS, turning the DSP into a customer-acquisition channel for commerce advertisers. The modeling required to do this well is the frontier we are investing in.

Responsibilities:

1、We are standing up a new Applied Science research team in Beijing to push that frontier. You will be the site lead and the technical and people leader for this team. You will build the team from the ground up, with full backing in terms of budget, headcount, and strategic resources to attract top-tier talent, set its research agenda, and own the path from research to live impact in the bidder.

2、Research agenda — commerce-DSP intelligence. Define and drive the applied-science roadmap across the full set of commerce-advertising problems, including but not limited to:

(1)Bidding & auction theory — bid optimization, mechanism design, budget pacing, value/ROAS-based bidding

(2)Conversion & ROAS prediction — large-scale CVR and value models under delayed, sparse feedback

(3)Contextual bandits & sequential decisioning — exploration/exploitation in live bidding

Embeddings & recommendation — user / item / creative representations, retrieval and ranking for commerce intent

(4)Incrementality & causal inference — measuring true lift, experiment design, attribution

Research-to-production. Own the line from idea to model running in the live bidder — this is not a hand-off-and-forget research lab. Partner with Bangalore platform and serving teams to ship models that hold up at millions of QPS and sub-50ms latency.

3、Build & lead the Beijing team. Hire, grow, and lead a high-caliber applied-science team (initial shape below). Set the bar for research rigor and engineering quality, establish the team's operating cadence and culture, and own its day-to-day technical and people leadership as site lead.

4、Cross-site partnership. Operate as one org with Bangalore — joint roadmap, shared standards, clear ownership boundaries, and tight collaboration with product and engineering leadership across sites.

Requirements:

1、15+ years or equivalent experience in applied science / machine learning, with several years building and leading technical teams.

2、Demonstrated track record shipping ML systems to production at scale — models running in live, latency-sensitive systems, with demonstrated production impact.

3、Deep expertise in one or more (ideally several) of: auction / bidding / mechanism design, large-scale prediction (CTR / CVR / ROAS), recommendation / embeddings, causal inference / incrementality.

4、Strong command of modern deep neural network architectures applied to this problem space — e.g., deep CTR / CVR networks, sequence and transformer models for user-behavior modeling, two-tower retrieval, and multi-task learning for ranking and prediction

5、Reinforcement learning and contextual bandits for sequential decisioning in the auction — e.g., RL for bid optimization and budget pacing, and exploration/exploitation in live bidding

6、Adtech / DSP / RTB / programmatic / real-time bidding background

Commerce, retail media, or performance-marketing experience (optimizing to purchase / ROAS)

7、Strong hands-on technical credibility — able to set direction, review research, and unblock hard modeling problems, not just manage

8、Proven ability to hire, grow, and retain senior applied scientists and engineers

Excellent communication and cross-site collaboration; fluent English

9、PhD preferred in a quantitative discipline, including CS, ML, Mathematics, Statistics, or Operations Research