AI- Full Stack Developer

What youll actually do

• Write production code. Repeatable work becomes compiled, testable binaries (Go-first); prompts drive persona behavior only. You build the tooling, not slideware about it.
• Operate agent fleets. Take a decomposed backlog, fan out parallel streams on conflict-free work, and bring it back through review gates.
• Build security in, not on. TDD test-writers ahead of coders, a dedicated validation stage (models hallucinate passing tests), persona judging against threat/security models, failures auto-looped as issues. Signed builds, SBOMs, and access control are table stakes.
• Integrate natively. Azure + GitHub Copilot, build-what’s-missing in-stack. No third-party platforms, no AWS.
• Run model-agnostic. Personas across Claude / Copilot / ChatGPT because they fail differently; disagreement is signal. Manage token and context economics as an engineering constraint — just-in-time context loading, per-task budgets, alerts.
• Own your slice end to end. Design, develop, test, integrate, deploy, document, and KT — across the full SDLC.

Must-have

• Senior engineer who still ships daily — hands-on-keyboard delivery, no exceptions.
• Deep Azure + GitHub Copilot delivery experience (primary stack, not a footnote).
• Strong Go (or equivalent systems language) for deterministic, testable orchestration tooling.
• Hands-on experience operating multi-agent / multi-model workflows in real delivery, not demos.
• Security-native instincts — TDD discipline, secure SDLC, threat modeling; comfort in a vulnerability-response context.
• Full-SDLC ownership of a work stream: design through deployment, docs, and knowledge transfer.

Nice-to-have

• SIEM / SOAR / EDR, identity, vuln-management, or CMDB / ITSM integration experience.
• Eval-harness, red-team, or agent-guardrail work (blast-radius sandboxing, reversibility, confidence thresholds, audit trails).