Senior Python Engineer (part-time) — AI-assisted development reviewer / Spanish-speaking

We're a small legal-tech effort in the Dominican Republic building an internal case-management platform for mortgage enforcement and litigation workflows. The product is designed in detail by our director (product + legal) and most of the code is written by an AI coding agent (Claude Code). We're not hiring someone to type code in a corner — we're hiring an engineer to direct that AI agent, review its output with real engineering judgment, and keep the codebase clean as it grows. This is a part-time, ongoing role. Spanish communication is required; the codebase and domain are in Spanish. Direct an AI coding agent to implement features that are already specified (specs are handed to you — you don't design them). Review the AI's output against the approved spec before it goes to the director for final approval. You're the first line of review: does this do what the spec says and nothing more? Does it reuse what already exists? Is it one concern per PR? Catch the drift AI agents are prone to: unrelated changes bundled into one PR, reinvented code instead of reuse, scope creep, mislabeled things. Fix bugs and handle the unglamorous work: migrations (forward-only), deployment ops, smoke checks. Maintain architectural continuity over time — no duplication, no dead code, no silent divergence from the design. You're a strong fit if you Have solid Python + SQL (SQLite, transactions) and a real Git/PR workflow. Can read a diff and smell what's wrong — not just "the tests pass." Are comfortable directing and distrusting generative AI: you verify, you don't rubber-stamp. Practice disciplined testing, idempotency, forward-only migrations. Communicate clearly in Spanish and can say "this went out of scope" without drama. Nice to have Streamlit experience (or willingness to learn its session_state/rerun quirks quickly). Experience in regulated domains — legal or financial — with sensitive-data handling. Hands-on with Claude Code / Copilot / Cursor or another coding agent. Not part of this role Legal or product decisions — those belong to the director. Designing features from scratch — they come specified. Stack Python 3.11 · SQLite (WAL) · Streamlit · pytest · Anthropic SDK · GitHub (squash PRs, CI with ruff + pytest + gitleaks) · Linux deploy (systemd + Caddy). Engagement [Part-time, ~15–20 hrs/week to start], ongoing, with room to scale based on fit. [Time zone: overlap with America/Santo Domingo (GMT-4) preferred]. We start with a paid 2-week trial on bounded, non-critical work before any longer commitment. [Rate: your bid]. How to apply (please read — this filters applicants) In your proposal, answer this short scenario in 3–5 sentences: "You're handed a PR titled 'fix bug X', but the diff touches 8 files, adds a brand-new search feature, and changes an unrelated module. All tests pass. What do you do, and why?" Proposals that just paste a generic CV without answering this will be skipped. Bonus if you briefly describe how you review AI-generated code without rubber-stamping it. 