Director of Engineering

Cardiff's edge is how quickly we can convert a lead into a funded customer. Today our average turnaround is 24 business hours; our goal is minutes, through automated systems with human-in-the-loop oversight. We have the projects, the budget, and the stakeholders lined up to get there. What we need is one person accountable for the whole department: for what ships, when it ships, and whether it holds up in production.

This is the senior technology role at Cardiff, and it is a hands-on job rather than a strategy seat. You will run the engineering department day to day, carry the delivery dates yourself, and build alongside the team when that's the fastest way to unblock them.

We are all in on AI, and we mean it structurally rather than as a tool choice.

Software factory. We're building an engineering organization where agents do the mechanical work of software; specification, implementation, testing, review, and deployment run as orchestrated loops, and engineers spend their time on judgment, architecture, and correctness. We want the department's output measured in systems shipped per month, not story points per sprint.

Agents end to end. Our production agents already talk to customers over SMS and email, work renewals and win-back campaigns, and take internal workflows off people's plates. The next step is agents that carry a process from trigger to funded outcome, with the loops, tool use, evals, and human checkpoints to be trusted with it.

Knowledge graph and total intelligence. Cardiff's institutional knowledge is scattered across Salesforce, Snowflake, email, Slack, Zoom, and documents. We're consolidating it into a queryable graph so that any agent or any person can get the full picture of a merchant, a deal, or a decision in one call. Every system you build should feed that graph rather than create another island.

Adopt before you build. Where the work isn't core IP, a strong open-source project beats owning the support and upgrade burden of a homegrown clone; Buzz, the agent-native workspace Block open-sourced, is the kind of thing we'd rather adopt and extend than rebuild.

If that philosophy is one you've already been living, this role is a place to run it at full speed with real budget behind it. If it reads as aspirational to you, this isn't the right seat.

You own Cardiff's engineering department: the in-house engineering team, our outside contractors and offshore Salesforce developers, and the vendor relationships our roadmap depends on. You report to the Managing Partner and sit on the leadership team alongside operations, sales, underwriting, marketing, and finance; those departments are your internal customers.

You'll set the delivery cadence, decide what gets built and what gets bought, keep production healthy, and be the single person the business goes to for a straight answer on status. The team is small and senior. Expect roughly 60% leading (planning, prioritization, review, unblocking, hiring) and 40% building; the ratio shifts toward building when we're behind.

Cross-department program management runs in parallel to you rather than under you, and underwriting decisioning is owned by a dedicated hire who works directly with the Underwriting department. You'll partner with both; you won't be carrying their calendars.

Delivery. Every committed date in the department, including the ones you inherit

The team. Hiring, onboarding, one-on-ones, performance, and growth for in-house engineers and contractors; you'll inherit an open Senior Full-Stack Engineer (AI-Native) req and run that search

The platform. Our Salesforce/Fundingo core, the Cardiff API, Snowflake, cloud infrastructure, and the migration path off legacy systems

AI products in production. Customer-facing agents for collections, renewals, and win-back campaigns, plus internal agents that take work off people's plates

The knowledge layer. The graph and the pipelines that feed it, and the standard that new systems connect to it

Vendors. Technical evaluation, integration ownership, and honest reporting on whether a vendor is earning its invoice

Production quality and security. Uptime, data integrity, access control, and the incident response when something breaks

First 30 days: You've met every stakeholder, inventoried what's in flight, and given the Managing Partner a written assessment of each project: real status, real risk, and a date you're willing to be held to. You've told us at least one thing we're doing that you'd stop.

First 90 days: A weekly delivery rhythm the business can see without asking for it. At least two stalled projects either shipped to production or formally killed. One agent-built workflow running end to end with the evals and human checkpoints to prove it's safe. Your first hire in process.

Measurable movement on time-to-fund, with the automation to show for it

A team that ships on a predictable cadence, where the engineers are visibly better than when they joined

The knowledge graph in real use, with agents and people querying it instead of hunting through five systems

A roadmap the business trusts, because the last three quarters of commitments landed

Systems that stay up, and a technology spend you can defend line by line

Run planning and prioritization with the department heads, and say no when the answer is no

Give proactive status: the business hears about slipped dates from you, on the day you know, not from asking

Design and operate agent systems: orchestration loops, multi-step tool use, retrieval against the graph, evals, and the human checkpoints that make them safe to run on customers

Review code and architecture, and stay close enough to production to know when an AI-generated pull request is wrong

Manage the offshore and contract developers to the same standard as the in-house team

Build against our stack: Salesforce, Snowflake, Gmail, Zoom, Slack, Twilio, and internal tools

Make AI tooling the department's default way of building, and hold the line on reading what it produces before it ships

Ship a working v1 today and improve it Monday, instead of a "perfect" v1 a month from now

Own the incident when something breaks in production, including the write-up afterward

8+ years building production software, with 3+ years managing engineers directly; you've run a team of 4 or more

A track record of hitting dates on projects you owned, and a clear story about a date you missed and what you did about it

Still technical: you've written and shipped production code in the last year, and you can read a pull request and catch what's wrong with it

Agent systems in production, not prototypes: orchestration loops, graphs, multi-step tool use, multi-agent patterns, RAG over real data, and evals you actually run

Coding agents as your primary way of building; Claude Code, Cursor, Codex, or similar, with a recent example of delivering something 3–5x faster than you could have without them

Range across the stack: Python, modern JavaScript/TypeScript, SQL, REST/GraphQL, and cloud deployment (AWS preferred)

Build-vs-adopt judgment: fluent enough in the open-source AI ecosystem to reach for a strong community project rather than hand-rolling what already exists

Communication discipline: you escalate early in writing, and your stakeholders always know where things stand

Nice to have:

Knowledge graphs, entity resolution, or unified data layers built for machine consumption

Fintech, lending, payments, or another regulated data environment

Salesforce APIs, Snowflake, or replacing a CRM-as-application architecture

Contributions to open-source AI projects

Experience as the most senior technical person in a small company

We weigh what you've shipped in the last 90 days more heavily than the brand on your resume. A builder with a public track record of working agent systems beats a longer tenure that stopped at experimentation.

You ship: a rough v1 in production beats a polished v1 in staging. You're fine with priorities that move. When something is blocked, you say so the same day, in writing, with what you need.

You know the frontier better than we do. We're not looking to teach you the latest capabilities; we're looking to be wowed, and to learn from you.

You want ownership, and you understand what comes with it. You'll have direct access to the Managing Partner, a real budget, and the authority to make technology decisions. You'll also be the person who answers for the outcome.

Want a strategy or advisory seat rather than accountability for delivery

Have been out of the code long enough that you'd need someone to translate a pull request for you

Prefer structured corporate environments with clear lanes and layers between you and the business

Solve problems offline and come back with results, instead of communicating while you work

Use AI tools to assist your old workflow, rather than having rebuilt your workflow around them

Cardiff lends to small businesses. Since 2004 we've funded more than $12 billion across the country, to businesses in logistics, healthcare, construction, professional services, and plenty in between. We're building our own underwriting technology, and we still answer the phone; our customers need capital fast, and they want a straight answer from a person who can actually help.