Forward-Deployed AI Engineer

AHL + Saaf AI - Mortgage Lending, Reimagined Saaf AI is building the future of mortgage lending by combining cutting-edge AI with proven lending operations. Saaf AI is part of American Heritage Lending, a top-10 private lender processing billions in loan volume, with 15+ years of growth. We are backed by some of the largest asset managers and funds and are growing fast. We're not experimenting with AI. We're deploying it. From underwriting to document processing to borrower experience, all are shaped by AI. If you want to work somewhere that uses AI as a core building material - you're in the right place. The Role We're hiring a Forward-Deployed AI Engineer to design and ship AI-powered products and agents from end to end. This isn't an ML research role or a prompt-engineering sandbox. It's a hands-on builder role - you own the full arc from "here's a business problem" to "here's a working AI system in production." Mortgage lending is one of the last industries to be meaningfully transformed by AI. The problems are hard, the data is rich, and the stakes are real. You'll work directly on the things that matter most: automating underwriting decisions, building intelligent borrower-facing experiences, creating agents that navigate complex loan workflows, and using AI to compress tasks that currently take days into minutes. This role is right for you if: • You've shipped AI-powered products or agents to production - not just prototypes • You think in systems: LLM calls, tool use, retrieval, evaluation, and feedback loops are design primitives to you • You move fast and take ownership - you'd rather build the thing than debate the spec • You want to be in a market where AI genuinely changes lives, not just optimizes click-through rates What You'll Build AI Agents & Autonomous Workflows Mortgage lending is a process-heavy industry - condition management, document review, borrower communication, underwriting analysis. You'll design and ship agents that automate or accelerate these workflows end to end: • Build multi-step LLM agents that can navigate loan files, evaluate conditions, flag exceptions, and surface recommendations to underwriters • Design agentic workflows with tool use, structured output, memory, and human-in-the-loop checkpoints - not one-shot prompts • Integrate AI agents with internal systems, external data sources, and third-party APIs to close the loop between insight and action AI-Powered Document & Data Intelligence Mortgage processing is document-intensive. You'll build the AI layer that makes documents machine-readable and actionable: • Extend and improve document extraction pipelines - income documents, appraisals, title, and more - using the latest extraction and classification models • Build retrieval-augmented systems that let underwriters and borrowers query complex loan files in natural language • Design feedback mechanisms that use reviewer corrections to improve extraction accuracy over time Borrower & Loan Officer Experience You'll use AI to make the borrower journey faster, clearer, and less stressful - and give loan officers AI-powered tools that make them dramatically more effective: • Build AI-generated task lists and smart checklists that adapt to each borrower's loan scenario • Create intelligent routing and prioritization - which loan needs attention now, which borrower is at risk of drop-off, which condition is the blocker • Design conversational AI experiences that guide borrowers through complex decisions without requiring a loan officer on every interaction AI Evaluation & Quality Systems You'll care about whether the AI actually works - not just whether it runs: • Build evaluation frameworks that measure AI output quality against ground truth, not just latency • Design human review workflows where AI confidence is low, and use those reviews to create labeled training data • Monitor production AI performance - accuracy, coverage, failure modes - and iterate relentlessly What We're Looking For Must-Have • AI systems builder: You've shipped LLM-powered features or agents to production. You've dealt with the real problems - latency, hallucination, evaluation, prompt regression, retrieval quality - not just the demos. • Strong Python: You write clean, production-quality Python. You're comfortable with async, APIs, data structures, and the engineering discipline that real systems require. • LLM tooling fluency: You've worked with the major LLM APIs and at least one agent orchestration framework. You know what structured output, tool calling, RAG, and evals look like in practice. • Product thinking: You can go from "here's a business process" to "here's an AI system design" without needing someone to translate. You think about the user, not just the model. • Ownership without guardrails: You drive features end to end - requirements, design, implementation, deployment, monitoring. You don't hand off to other teams at the hard parts. • Fast i...