AI Refactoring Manager

hivefinancialsystems; Atlanta (GA); On-site; USD 150,000 - 210,000; Full time

Job summary

Hive Financial Systems in Atlanta is seeking an AI Refactoring Manager to lead the rebuild of our legacy lending platform onto a new multi-portfolio architecture. You will guide AI-driven code generation while owning judgment, review, and quality, and you will hands‑on read the code.

This leadership role requires 8+ years in software engineering, with proven experience in agentic AI workflows, and deep knowledge of C#/.NET, SQL Server, CI/CD, and cloud deployments (Azure).

Qualifications

• 8+ years in software engineering, including 3+ years leading engineers or a technical program.
• Led at least one large legacy modernization through to production.
• Proven experience in agentic AI engineering and workflow creation for production work.

Responsibilities

• Own the rebuild roadmap: sequence legacy capabilities to migrate first based on risk and value.
• Design and run the AI-assisted engineering workflow for code generation and handoff to humans.
• Build and maintain prompts, agent instructions, and reusable AI skills according to architecture standards.
• Enforce test-first discipline; ensure AI-generated code passes acceptance tests.
• Review AI-generated code and set the human review bar, focusing on money math, borrower data, and portfolio isolation.
• Separate live code from dead code and document legacy business rules.
• Collaborate with Compliance and Product to capture regulated behavior in tests.
• Track delivery health and AI effectiveness metrics; report and optimize.

Skills

C#/.NET

AI engineering

CI/CD

Azure

Kubernetes

Terraform

Helm

Bicep

Argo CD

Code review

Tools

Kubernetes

Helm

Terraform

Bicep

Argo CD

GitOps

Job description

Location: Atlanta, Georgia (Buckhead)

Employment Type: Full-time, Exempt

Department: Engineering

About Us

Hive Financial Systems is a leading fintech company transforming financial inclusion through AI-driven lending technology. We power 3 active lending partners serving 70,000+ active consumers, building the infrastructure that enables responsible lending to underserved markets. Our team is small, technical, and moving fast-and AI is central to how we do it.

Role Overview

We are looking for an AI Refactoring Manager to lead the rebuild of our legacy lending platform onto a new multi-portfolio architecture, with AI coding agents doing much of the code generation and people owning the judgment, review, and quality bar. This is a hands‑on leadership role: you will set the method, run the work, and still read the code.

Key Responsibilities

• Own the rebuild roadmap: sequence which legacy capabilities move first based on business risk, dependencies, and value.
• Design and run the AI-assisted engineering workflow: how agents read legacy code, extract business rules, draft services, and hand off to human review.
• Build and maintain the prompts, agent instructions, and reusable skills that encode Hive's architecture standards so AI output starts close to right.
• Enforce a test‑first discipline: failing acceptance tests exist before code is written, and AI-generated code passes the same gates as human code.
• Personally review AI-generated code and set the review bar for the team, with particular attention to money math, borrower data, and portfolio isolation.
• Separate live code from dead code in the legacy estate and document the business rules buried in it before they are lost.
• Partner with Compliance and Product so regulated behavior (disclosures, payments, adverse action, consent) is captured and proven in tests.
• Track and report delivery health (lead time, deployment frequency, change failure rate, recovery time) and AI effectiveness (rework rate, defects traced to AI output).
• Coach engineers on working well with AI agents, and decide where AI should not be used.
• Evaluate AI coding tools and models, and manage their cost, access, and security posture.

What Success Looks Like in the First 12 Months

• By Month 3: A documented, repeatable AI‑assisted rebuild method is in daily use by every rebuild team.
• Per service: Legacy business rules and acceptance criteria are extracted and signed off by Product before rebuild begins.
• Every release: Rebuilt services meet 90%+ line and branch test coverage, with failing tests written before code.
• By Months 6 and 12: Lead time and deployment frequency improve quarter over quarter without a rise in change failure rate.
• By Month 12: An agreed set of legacy capabilities is live on the new platform with legacy equivalents switched off, and zero compliance or data‑isolation defects from AI-generated code have reached production.

Qualifications

• 8+ years in software engineering, including 3+ years leading engineers or a technical program.
• Has led at least one large legacy modernization (monolith to microservices, re‑platform, or rewrite) through to production.
• Proven experience in agentic AI engineering and workflow creation: designing, building, and running multi‑step AI agent workflows (e.g., with Claude Code or similar agent frameworks) on production work, with a clear view of where they fail.
• Strong in C#/.NET; comfortable reading older .NET and SQL Server code.
• Practices test‑driven development and can explain why coverage numbers alone are not quality.
• Working knowledge of CI/CD, trunk‑based development, and cloud deployment (Azure preferred).

Preferred Skills

• Experience in lending, payments, or another regulated financial domain (TILA/Reg Z, NACHA, FCRA, GLBA).
• Kubernetes, Helm, Terraform or Bicep, and GitOps tooling such as Argo CD.
• Has written agent instructions, prompt libraries, or internal AI tooling used by a team.
• Multi‑tenant system design where data isolation between tenants is a hard requirement.
• Familiarity with DORA metrics and observability tools such as DataDog.

Competencies We Hire For

• Agentic Engineering Judgment: Knows when to trust, verify, or reject agent output, and can show agents they built and the failures they caught.
• Workflow Design: Breaks complex engineering work into agent workflows with clear steps, checkpoints, and human review p