Founding Applied AI/ML Engineer

The role

You would be the engineering owner of the platform that everything else stands on. There is a working system in production with paying customers and real documents flowing through it. Your job is to take ownership of it, harden it, and scale it, working directly with the founder and a small in-house team.

Three ideas define this job. First, accuracy is the product. A bank does not buy a demo, it buys the confidence that every number an underwriter sees traces to the correct line of the correct source document. The evaluation harness that proves that is not a side project here. It is the moat, and you own it.

Second, cost per page is the margin. We price by the deal and we pay by the page. Model routing, caching, and fine-tuning smaller models where they earn their keep: the gap between what inference costs and what a deal pays is a curve you own end to end.

Third, the data compounds. Every deal processed makes the repository deeper and the extraction smarter. Designing the golden datasets and pipelines that turn daily volume into a durable advantage is the long game, and it starts now.

What you will own

• The document intelligence pipeline: extraction across appraisals, rent rolls, tax returns, leases, and loan documents, from ingestion and OCR through LLM orchestration to structured, verifiable output.
• Evaluation and accuracy: eval harnesses, golden datasets, and regression suites for LLM-backed extraction. An extraction that cannot be verified does not ship.
• Data engineering: the pipelines, schemas, and internal repository that turn thousands of processed documents into the dataset this company's future is built on.
• The unit economics of inference: cost per page at the current model mix, and the roadmap that brings it down without giving accuracy back.
• Production engineering: reliability, observability, and the security posture our SOC 2 obligations and bank diligence reviews demand.

Who this is for

• Five or more years building production ML and AI systems that real users depend on. Shipped, operated, and debugged in production, not just modeled in a notebook.
• You have built LLM-backed systems with your own hands: retrieval, multi-step pipelines, structured extraction, prompt and model optimization against a measurable target.
• You distrust your own pipeline until the evals say otherwise, and you know why PDFs are harder than they look.
• Strong data engineering fundamentals: pipelines you have owned at scale, fluent SQL, sane orchestration.
• AI-native as a daily habit. You code with frontier AI tools and can show us the workflow, not just name the tools.
• Helpful but not required: lending, fintech, or commercial real estate exposure; time in a SOC 2 or otherwise audited environment.

What we offer

Competitive base salary plus a meaningful ownership stake in the outcome, at founding-hire scale. We put real numbers on the table in the first conversation, not the last. Direct lines to the CEO, the credit authority, and the customers. Your code reaches production, and paying customers, the week it is ready. Dallas based, on-site.

How to apply

Apply below with your resume, a link to something you built if one exists, and an answer to one question in a few sentences: tell us about a time an AI or ML system you built was confidently wrong in production. How did you catch it, and what did you change so it could never again fail silently?

LenderBox is an equal opportunity employer. If you need an accommodation at any point in this process, tell us and we will arrange it.