COBOL Forensics & Modernization Engine
COMP-3 Boundary Discovery — Public synthetic probe: $40,812.81, clone & run
The Problem
Enterprise mainframe systems (40-year COBOL codebases) contain undocumented COMP-3 packed-decimal boundary conditions that cause silent data corruption in production. Traditional auditors lack the tooling to detect these — they require manual byte-level forensic analysis.
Approach
Built automated COMP-3 / PICTURE-clause boundary analysis toolchain. Adversarial audit methodology: inject boundary conditions at field level, trace propagation through batch processing chains, surface any silent truncation or overflow. The $2.3M find is a production engagement under NDA. The same failure class is reproduced publicly on a synthetic, fixed-seed corpus — clone and run cobol-pic-probe ($40,812.81 compounding loss, 47,312 records, 31 pytest assertions, no client data).
Proof Metrics
Want this on your team?
This work maps to 4 qualifying roles. Remote, Q3 2026.