Meridial Marketplace, by Invisible · 11 hours ago
COBOL Expert (SME) - Freelance Project
Meridial Marketplace, by Invisible, is seeking a highly experienced COBOL professional to serve as a consultant on AI training data projects for leading AI model builders and enterprises. The role involves defining success criteria, reviewing outputs, and providing targeted guidance to improve quality and speed, ensuring the successful delivery of domain-specific annotated datasets and code samples.
Computer Software
Responsibilities
Define COBOL-specific quality success metrics for code annotation and dataset labeling projects
Develop project-specific SOPs, QA rubrics, and reference materials to ensure outputs align with client technical standards
Review project outputs (COBOL scripts, code annotations, legacy modernization samples) against defined standards, flagging and correcting defects before client delivery
Conduct structured QA passes on deliverables; track, flag, and resolve defects efficiently to hit delivery timelines
Return work to contractors with precise remediation notes and context on COBOL syntax, logic, and legacy system patterns
Provide advisory input on tools, frameworks, emulators, and workflow improvements to maintain quality benchmarks in mainframe and batch-processing environments
Handle spec changes and edge-case scenarios (e.g., different COBOL dialects, EBCDIC vs ASCII encoding, JCL dependencies) and draft corresponding acceptance criteria or workarounds
Curate libraries of “gold standard” COBOL code samples, modernization examples, and dataset annotations for calibration and cross-project consistency
Participate in technical assessments of contractor talent, including reviewing COBOL code assessments and task-based evaluations
Review sample outputs from contractors and provide clear, actionable written feedback to improve code correctness, readability, and efficiency
Develop targeted training and calibration resources, such as:
COBOL code quality guidelines (e.g., data division consistency, paragraph structuring)
Best practices for clean, maintainable procedural code
Reference documentation for legacy system interaction patterns
Dataset labeling standards for COBOL-related model training
Advise on technical scoping and requirements during project setup, including COBOL versioning, JCL integration, and mainframe data formats
Provide expert guidance for edge cases and spec changes, such as handling copybooks, variable-length records, or integration with DB2 and VSAM
Contribute to post-project reviews to capture lessons learned and continuously refine standards
Identify and summarize client system insights, such as recurring syntax issues, logic errors, or data formatting inconsistencies
Build dashboards or defect trackers with categorized issues to surface recurring themes and drive process improvements
Conduct post-mortems to analyze defect trends and propose updated QA steps, documentation improvements, or refresher training
Qualification
Required
5+ years of professional experience in COBOL programming, mainframe systems, or enterprise application maintenance, with demonstrable project impact
Mastery of COBOL
Proven ability to define, enforce, and maintain high technical standards in legacy system environments
Strong communication skills, with the ability to convey technical feedback clearly to both engineers and non-technical reviewers
Experience developing technical documentation, QA frameworks, or training content for legacy or mainframe systems
Ability to deliver against fixed project timelines and scopes
Exceptional attention to detail and a commitment to accuracy, consistency, and documentation discipline
Fluency in spoken and written English with clear, concise technical writing
Preferred
Familiarity with JCL, DB2, CICS, or VSAM
Company
Meridial Marketplace, by Invisible
We are the AI training and scaling partner for the leading foundation model providers, enterprises, and governments, bridging the gap between AI potential and production.
Funding
Current Stage
Growth StageCompany data provided by crunchbase