Temporal Grounding for AI: Building Calendar Systems Worth Trusting

Date errors in AI-generated outputs are particularly insidious because language models can confidently wrap wrong answers in plausible reasoning. Project Parva addresses this by building temporal grounding infrastructure for systems requiring accurate Bikram Sambat conversion and complex calendar logic—essential for fiscal compliance, payroll, and audit trails in Nepal-based operations. The project demonstrates why calendars with legal or operational significance must be treated as infrastructure, not utilities.