Hi! I’m Zach! I’m a software engineer in the New York City area. I’ve worked in different roles across very different organizations, but I’ve spent the most time working on projects in the data and finance spaces. I highly value code quality and I view data the same way I do code: it should be clean, well tested, and well documented.
I work hard to make the teams I’m on healthy, collaborative, and fun. I think teams with high psychological safety, morale, and mutual respect will work and communicate as effectively possible to create the highest quality software while maintaining a sustainable pace. I believe heavily in the eXtreme Programming methodology and practice pair programming, test driven development, user centered design, and continuous integration (and improvement) whenever possible.
☎️ Contact
📧 [email protected]
🔗www.linkedin.com/in/zach-marcin
👨🔧 Work experience
Staff Software Engineer
Mechanical Orchard
(Jan 2024 – May 2026) New York, NY — Remote
Mechanical Orchard is a mainframe modernization company that rewrites critical mainframe applications into cloud-native systems using its AI-native Imogen platform — capturing live behavioral specifications from running mainframes and recreating them incrementally with continuous behavioral equivalence verification. Worked across multiple client engagements modernizing CICS and batch workloads, and led the training programs that enable engineers to do this work at scale.
Responsibilities
- Replaced dozens of legacy CICS terminal screens with modern Phoenix LiveView web interfaces; conducted user research with operators that resulted in zero-retraining adoption and a strong stated preference for the new UI over the existing CICS terminal
- Built a bespoke behavioral-equivalence testing harness that drove the legacy CICS screens via a web-based terminal emulator and ran the equivalent flows through the new web app, comparing outputs step-by-step to guarantee behavioral parity throughout the migration
- Created digital twins of mainframe batch jobs — Elixir and Java re-implementations that produced byte-identical output to the original COBOL given the same input — driving the LLM-assisted (Claude, ChatGPT) Java generation pipeline alongside Mechanical Orchard's Imogen tooling, reviewing and correcting model output to ship code that passed continuous equivalence checks
- Designed and delivered the end-to-end training program for Imogen: owned the curriculum, built mock mainframe environments, authored 30+ domain-specific COBOL and JCL programs, and developed separate tracks for hands-on engineers, PMs, and leadership — training 20+ engineers across client engagements and bespoke sessions for other MO teams
- Built an LLM-powered self-guided training environment on Coder with an Airflow-based mock mainframe scheduler and custom DAGs; designed an LLM tutoring "skill" that walked trainees through progressively harder modernization jobs, answered questions without giving away solutions, and graded submitted code — including coaching trainees on how to use LLMs effectively for modernization work
- Shaped Imogen's product direction by feeding trainee and client-user feedback back into UX simplifications and partnering with sales to refine the demo experience; mentored teammates daily through pair programming in an XP-style team
Technologies Used
Elixir, Phoenix LiveView, Java, COBOL, JCL, Python, z/OS, CICS, DB2, Postgres, Apache Airflow, Coder, Claude, ChatGPT, Kubernetes, Terraform, Docker, Nix