IBM Software Developer - WatsonX Code Assistant for Z
Software Developer at IBM building AI tooling for WatsonX Code Assistant for Z - enabling mainframe code understanding and translation using LLMs.
A 12-month internship at IBM (May 2025 - April 2026) building AI tooling for WatsonX Code Assistant for Z, enabling enterprise mainframe code understanding and modernization.
The Challenge
Mainframe systems running COBOL and PL/1 power critical infrastructure across financial services, healthcare, and government. These systems process:
- 95% of ATM transactions
- 80% of in-person transactions
- Trillions of dollars daily in financial transactions
The workforce maintaining these systems is aging, and modern developers need AI-powered tools to understand and modernize this critical code.
My Contributions
Model Benchmark Pipeline
Designed and implemented a benchmark pipeline evaluating Claude, Llama, Mistral, and internal IBM models for code understanding and translation tasks. Results directly influenced product model selection decisions.
Data Dictionary Feature (Shipped to Production)
Built the data dictionary integration that enriches LLM context with variable definitions, types, and usage patterns extracted from legacy code. This enables significantly higher-quality code explanations by giving the model semantic understanding of the codebase.
VSCode Extension - A2A & MCP Integration
Extended the VSCode extension with Agent-to-Agent (A2A) protocol and Model Context Protocol (MCP) integrations, connecting IBM's existing profiling and analysis tools to AI workflows. This enables richer, more accurate code explanations by leveraging programmatic analysis alongside LLM reasoning.
Leadership - Future Blue Intern Council
Led IBM's Future Blue intern council as President, organizing executive networking events attended by IBM Canada's President and senior VPs. Built community across the intern cohort through social events and career-focused programming.
Technologies
- Legacy Languages: COBOL, PL/1
- Modern Languages: Python, Java
- AI/ML: LLM integration, AI Agents, MCP protocol, A2A protocol
- Developer Tools: VSCode Extensions, benchmark pipelines
Impact
This work directly contributes to IBM's WatsonX Code Assistant for Z, helping enterprises understand, maintain, and modernize their mainframe systems while enabling the next generation of developers to work with legacy code effectively.