Business teams model everything in Excel – pricing logic, risk calculations, financial forecasts. Engineers then rewrite that logic in Java. Every time. This translation process causes bugs, delays, and drift between "what the business thinks the logic is" vs "what's actually running."
HyperCell takes a different approach: load the Excel file, compile the formulas into a DAG, and execute them in-memory. No translation. The spreadsheet IS the code.
Technical details: - Parses Excel formulas using ANTLR4 - Builds a dependency graph for intelligent recalculation - 200+ functions (SUM, VLOOKUP, INDEX/MATCH, IF, NPV, IRR, etc.) - Cross-validated against Excel with 82,881 formulas at 100% accuracy - ~50ms to load a typical workbook, sub-millisecond recalculation
Use cases we've seen: - Insurance companies running rating models - Fintech doing real-time pricing - E-commerce with vendor-specific shipping rules
It's Apache 2.0 licensed. We extracted it from our production system where it's been running for 2+ years.
Happy to answer any questions about the architecture or implementation!