Django/Python Agentic Engineer

Agentic Engineer for Human in the Loop Code Review, including Django and RLS Permissions System Architectural Enforcement Remote | Ongoing Contract | Fixed Price per PR We are an education technology company shipping high volumes of AI-generated code through automated development pipelines. We need an experienced engineer to own human PR reviews, prevent regressions, coordinate fixes through coding agents, and keep releases moving. This is primarily a code-review and technical QA role, not feature development. Substantial fixes are delegated to agents; minor corrections can be made directly… This is a review role, not a build role. You'll audit the codebase at four milestones. At each one you'll assess architecture, security, and correctness, then give clear, prioritized findings. You'll review AI-generated code, so you need a sharp eye for subtle logic, security, and data-integrity problems that look fine on the surface. Occasional hands-on fixes may come up, but they are secondary. Must Have Skills - MUST HAVE Core Backend & Database: Extensive proficiency in Python, Django (ORM, migrations, sessions, testing), and PostgreSQL 16 (Row-Level Security/RLS, multi-tenant isolation, transactions, locking, indexing, and data integrity). Permissions & Security Architecture: Proven experience working with and understanding Django permission systems, django-guardian object grants, multi-tenant isolation, and database RLS. Understanding of customized django permissions systems is strongly preferred. Agentic Development Review: Hands-on experience directing, optimizing, and verifying AI-generated code (e.g., Claude Code, Codex), awareness of the common failure modes, and ability to spot overengineering and prevent techical debt from accumulating through automated and human in the loop spec driven development workflows, alongside the willingness and capability to work with agility and at speed, avoiding unnecessary ceremony without taking unnecessary risks. Critical Thinking: critically evaluating AI outputs, using your own judgment, within the context of our business goals, the law, and software development best practices, without relying solely on CI passes or AI explanations. Full Application Stack: Hands-on expertise with Django templates, HTMX, Alpine.js, Unfold, django-filter, django-tables2, and Crispy Forms. Engineering & AI Solutions: Track record of developing scalable AI solutions using robust Python coding, testing, version control, security practices, debugging, and sound design judgment. Self-Directed Execution: Demonstrated ability to quickly learn product requirements and repository conventions independently, consistently enforcing them across codebases. What you'll review We expect you to recognize common failures in AI-generated code: incorrect model assumptions, hallucinated fields, misleading tests, weakened assertions, incomplete refactoring, unrelated changes, duplicated logic, unnecessary abstractions, and regressions across concurrent PRs. You must also identify genuine Django and PostgreSQL problems: N+1 queries, race conditions, unsafe migrations, transaction errors, incorrect permission grants, RLS failures, tenant-data leakage, and security vulnerabilities. Example Deliverables (to be discussed and confirmed) Decision: Approve, Approve with Exception, Request Changes, or Escalate, tied to the reviewed commit. Findings: Exact file/line, confirmed defect, severity, affected functionality, and required correction. Verification: Tests and scripts executed, results, and browser QA evidence where relevant. Regression assessment: Existing functionality and dependencies potentially affected, including authorization and data integrity. Fix coordination: Hand substantive corrections to coding agents with clear instructions and acceptance criteria. Make trivial fixes directly where more efficient. Reverification: Review the agent's corrections, rerun relevant checks, and confirm the issue is resolved without introducing regressions. Exceptions and follow-ups: Document accepted noncritical failures and emerging technical debt requiring later attention. You own the review through resolution, not simply the identification of problems. Reviews must be concise, evidence-based, and actionable—not generic AI-generated commentary. Protect architectural coherence and spot technical debt early. Recognize anti-patterns before they spread across multiple agent-generated changes. Identify duplicate business logic, unjustified abstractions, accidental data duplication, incorrect framework use, N+1 access patterns, transactional weaknesses, unnecessary custom solutions, and increasing coupling. Consider interactions between PRs, database changes, permissions, deployment behavior, and related product features—not just whether one diff looks tidy. Explain the likely trajectory of a problem and recommend intervention at the point when the expected benefit justifies the work. Working expectations You must get up to speed quickly, investigate independently, direct coding agents effectively, and manage multiple concurrent PRs without constant supervision. We need someone who understands Django deeply but is pragmatic rather than rigid. Protect users, security, privacy, financial correctness, and critical functionality. Do not block delivery over stylistic preferences or low-impact issues. The architecture and engineering rules are established and must be followed. Protected authorization changes require CTO approval. You will be trusted to make routine review decisions independently, with clearly defined escalation boundaries. Payment and application Fixed payment per completed PR review, including coordinating and verifying necessary corrections, with agreed turnaround and volume expectations. Please provide relevant Django/Guardian experience, examples of significant defects you identified, experience using coding agents to resolve findings, your price per PR, and daily availability. Shortlisted candidates will complete a practical assessment involving an agent-generated Django PR, misleading tests, authorization and regression risks, agent-directed corrections, independent verification, and a justified merge decision. Starts part time and can go to full time. Must give your price for 10, 20, 30, 40 per week. Fixed cost and not hourly. You must also provide your resume in PDF. If you don't provide that information, there will be no consideration.