COBOL Test Workload Developer

1. About the Company

CloudFrame builds Continuum, the markets most capable deterministic COBOL-to-Java transformation platform, with guaranteed functional equivalence between the original COBOL and the generated Java.

What is less well known is how it works. Continuum has never transformed COBOL to Java in a single step. Under the covers it has always gone COBOL to optimized COBOL first, and then pattern-to-pattern into Java. That intermediate optimization stage has been running quietly inside every transformation we have delivered.

We are now opening that stage up to customers directly. Optimize in Place gives a customer access to the optimized COBOL itself a low-risk automation path that improves the performance and efficiency of their existing COBOL, on their existing mainframe, without a platform change and without a rewrite. Continuum also generates full documentation of the program and the changes applied. This is a mature component being surfaced, not a new product being invented. Our job now is to harden it for direct customer exposure.

2. The engagement

We need an experienced COBOL developer to build the test workloads we do not currently have. You will write deliberately inefficient COBOL the kind that genuinely exists in production at banks and insurers run it through Optimize, and tell us honestly whether the optimized COBOL that comes back is correct, measurably faster, and something a customers own development team would accept into their codebase.

You will validate the platforms generated documentation and its synthetic test data generation to the same standard. All three capabilities are linked: generated test data is what proves an optimization is equivalent, and the documentation is what gets the rewrite through a customers code review.

You are not being hired to demonstrate that the product works. You are being hired to find out where it doesnt, on our schedule and in our lab, rather than in a customers batch window.

3. What you will do

Design and build original COBOL test workloads representing classic performance anti-patterns, with the JCL, DB2 and VSAM setup and deterministic test data to run them end to end
For each case, also write the optimized version you would produce as a senior developer giving us a human expert benchmark to measure Optimize against, not just a pass or fail
Establish measured baselines on z/OS: CPU, zIIP-eligible CPU, and elapsed time
¢ Run cases through Optimize and assess the output on four axes: functional equivalence, measured performance gain, code quality and maintainability, and whether the change is safe given context the tool cannot see from the source
¢ Validate the generated documentation accuracy against what the code actually does, coverage of error and edge paths, and whether it explains the applied optimization well enough to survive a customer code review
¢ Validate Continuums synthetic test data generation structural validity against the copybook, referential consistency across files and tables, reproducibility, business plausibility, and above all whether the generated data actually exercises the branches, error paths and boundary conditions that a functional equivalence claim depends on
¢ Build adversarial test data specifically intended to break the functional equivalence claim: invalid packed decimal, boundary values, empty and single-record files, duplicate keys, SQLCODE edge paths, abend conditions
¢ Produce minimal reproducers for anything that breaks, so engineering can act on it the same day
¢ Grow the test inventory as you learn where the engine is weak, with the reasoning attached
4. Required skills
¢ 10+ years hands-on enterprise COBOL development in production mainframe environments
¢ Direct, personal COBOL performance and tuning experience you have reduced CPU or elapsed time on production programs and can prove it with measurement. This is the core qualification, not a bonus. If you have never owned a tuning exercise end to end, this is not the right role.
¢ Able to judge generated technical documentation whether a program description accurately reflects what the code actually does, covers the error and edge paths rather than just the main flow, and would let a developer unfamiliar with the program work on it safely
¢ Test data design for coverage able to construct datasets that drive every branch, error path and boundary condition in a program, and to judge whether generated data does the same
¢ Fluent JCL: multi-step jobs, conditional execution, GDGs, utilities, procedures
¢ DB2 for z/OS: embedded SQL, cursors, DCLGEN, BIND, EXPLAIN, reading access paths
¢ VSAM: KSDS, ESDS, RRDS, IDCAMS, buffer and AMP tuning
¢ Sequential and VSAM file handling, blocking, record formats
¢ COBOL compiler options and their runtime consequences OPT, ARCH, TRUNC, SSRANGE, TEST
¢ z/OS operational literacy: SMF data, job accounting, abend interpretation, batch window realities
5. Nice to have
¢ Language Environment runtime options and behaviour
¢ DFSORT or SyncSort, including exits
¢ Mainframe cost model familiarity: MLC, R4HA, Tailored Fit Pricing, zIIP eligibility
¢ Code review experience able to judge whether a rewrite is one a development team would actually accept
¢ Prior involvement in a modernization, migration or replatform program, from either side
¢ Java literacy sufficient to read transformed output
¢ Git and scripted, reproducible environment setup