Aadhaar (UIDAI): Biometrics at national scale

Context

India’s Aadhaar program operates biometric identity verification at unprecedented scale, requiring systems that are not only accurate but also operationally reliable, auditable, and fair across demographics and time. Research and production systems must handle long time gaps between enrollments, large population diversity, and strict evaluation standards, while remaining deployable in real-world conditions.

What we built

• Large-scale biometric de-duplication and face matching architectures designed for national identity verification.
• Age-robust face recognition approaches to improve matching reliability across long time gaps and changing appearances.
• Evaluation frameworks and benchmarking protocols to measure biometric performance at scale.
• Biochallenge: a public, nationwide benchmarking platform enabling standardized evaluation of biometric SDKs and algorithms.
• Research-to-production alignment, ensuring models and evaluation methods reflected real operational constraints.

Outcomes

• Improved face recognition performance, achieving a ~6% uplift at FMR@10,000 on large-scale benchmarks.
• Strengthened reliability of biometric verification across age variation and long enrollment gaps.
• Enabled transparent, repeatable evaluation of biometric systems through a public benchmarking initiative.
• Contributed to more robust, scalable, and operationally sound biometric identity systems.