Prateik Mahendra

Building High-Resolution Population Maps Without Compromising Privacy

Abstract:

High-resolution population mapping is becoming increasingly important for disaster response, public health, infrastructure planning, humanitarian logistics, and climate resilience. However, the same spatial precision that enables better decision-making can also increase the risks of inference, linkage, and re-identification. This talk explores how population intelligence can deliver actionable geographic insights while protecting individuals and vulnerable communities.The session presents a privacy-aware framework built around six stages: Observe, Represent, Infer, Protect, Validate, and Govern. This approach integrates Earth observation data, geospatial features, spatial learning, uncertainty estimation, privacy safeguards, fairness assessment, and lifecycle governance. The talk also examines why differential privacy alone may be insufficient for complex geospatial systems, highlighting complementary techniques such as federated learning, secure multi-party computation, trusted execution environments, synthetic population generation, and adaptive disclosure controls. Emerging AI capabilities—including vision and multimodal foundation models, graph neural networks, Bayesian deep learning, continual learning, and agentic AI—are explored for their potential to improve spatial reasoning, modeling accuracy, and workflow automation. Key challenges include geographic domain shift, uneven population representation, validation across heterogeneous data sources, computational reproducibility, and the inherent tension between increasing spatial resolution and preserving privacy. The session concludes with future directions such as privacy-aware foundation models, geospatial digital twins, causal spatial AI, autonomous agents, and trusted spatial intelligence platforms. These developments reinforce the central principle that privacy should not be treated as an afterthought, but as a core architectural requirement for trustworthy population intelligence.

Brief Profile:

Prateik Mahendra is a Data Engineer at Hinge Health with more than six years of experience in data engineering, analytics, privacy, and AI-powered automation. Previously at Meta, he led large-scale data infrastructure and privacy initiatives that delivered $14 million in OPEX savings, reduced regulatory audit turnaround time by 70%, and contributed to a 10% improvement in engineering efficiency that was adopted as a monthly executive KPI. His work spans cross-app soft-match detection, privacy cost measurement, graph-based data lineage, large-scale geospatial data pipelines, LLM-based automation, and AI-agent workflows for privacy and compliance use cases. Beyond his industry experience, Prateik is recognized for his contributions to AI, data privacy, and responsible data engineering. He has served as a judge for leading AI and technology competitions, contributed as a peer reviewer and session chair at an IEEE-sponsored conference, and delivered an invited talk on data privacy in the age of big data. He has also published research on AI-driven privacy and ethical data engineering, reflecting his broader contributions at the intersection of scalable data infrastructure, privacy engineering, and responsible AI.