Applied Scientist – User Intelligence

Overview of the Role:

We are looking for an Applied Scientist to develop and improve our user personalization intelligence through architectural and data research aimed at optimizing the efficacy of our targeting and pricing systems. More specifically, our team is exploring state-of-the-art attention-based architectures powered by multi-modal data, operating at InMobi's scale of trillions of events per day.

The Impact You'll Make:

• Build a pipeline of evidence-backed hypotheses and evaluate them through to a clear accept/reject decision, taking the winners to production
• Drive architectural research into attention and post-attention architectures while meeting the millisecond-level latency demanded by real-time advertising
• Advance our evaluation methodology across offline and online settings, ensuring model improvements translate into real business impact
• Leverage the breadth of data features across our product portfolio to improve our models and help guide product teams toward sourcing high-value data
• Contribute regularly to our internal model implementations, training pipelines, and documentation to help the team operate efficiently and share knowledge effectively
• Guide our infrastructure teams on medium- and long-term needs to increase our experimentation velocity
• Maintain an active research profile through external engagement — publishing, giving talks, and participating in top conferences

The Experience We Need:

• PhD in machine learning, data science, or a related field preferred; Master's degree also considered, particularly with a strong industry track record. Candidates nearing graduation are also welcome to apply.
• Demonstrated expertise in machine learning/deep learning, whether through peer-reviewed publications (e.g., NeurIPS, KDD, RecSys) or through industry experience building and shipping models that performed well at scale
• Fluency in the PyTorch or JAX ecosystem, with the ability to implement complex operations in vectorized form
• Expertise in deep recommender systems, including two-tower and attention-based architectures
• Comfort with large-scale distributed data systems and the ability to build production data pipelines for training and inference
• Understanding of the ads ecosystem (OpenRTB, SSPs, DSPs, ad exchanges, MMPs) is a plus