India’s retirement-income security architecture has undergone a fundamental structural transition over the past three decades, shifting from a defined-benefit (DB) logic centred on the Employees' Pension Scheme, 1995 (EPS-1995) toward a defined-contribution (DC) logic anchored in the National Pension System (NPS, introduced 2004), and, most recently, toward an explicitly hybrid design under the Unified Pension Scheme (UPS, 2025). This paper undertakes a secondary-data comparative analysis of defined-benefit and defined-contribution pension architecture in India, situating EPS-1995, NPS and UPS within the World Bank's multi-pillar pension taxonomy and against international pension-reform experience drawn from Chile, Sweden, the United Kingdom, Australia, and Central Europe. Drawing on EPFO and PFRDA administrative releases, OECD Pensions at a Glance indicators, and the academic and policy literature on pension privatization and its reversal, the paper finds that EPS-1995 exhibits classic vulnerabilities of an under-funded DB scheme a recorded actuarial deficit, a wage ceiling and minimum pension that have lagged wage growth, and costly ex post corrections such as the 2022 Supreme Court higher-pension judgment while the NPS, though fiscally robust and portable, leaves market, inflation and longevity risk almost entirely with the individual subscriber. International evidence, including Chile's 2025 reintroduction of an employer-financed defined-benefit layer into a pure-DC system and the politically costly reversal of mandatory private accounts in Hungary and Poland, suggests that durable old-age income security in India is unlikely to rest on a binary DB-versus-DC choice but on a deliberately layered, multi-pillar design. The paper closes with policy implications for EPFO, PFRDA and India's wider pension ecosystem.