AI Developer - Underwriting, Ops Automation & AI Dispatch (Freight)

OVERVIEW We're a fast-growing US freight factoring and fuel-card company serving small trucking fleets (1-10 trucks). We're building an AI-first platform to fund carriers faster, catch fraud earlier and run operations with a lean team. Experienced underwriters stay in the loop for final credit decisions; we need a hands-on AI developer to build the tooling that makes them faster and our operations close to touchless. WHAT YOU'LL BUILD - AI-assisted underwriting: ingest onboarding files (KYC, operating history, bank statements, invoices, UCC/lien searches), pre-score applicants, flag red flags and summarize each file for a human underwriter. - Fraud and invoice screening: detect double brokering, falsified paperwork and suspicious invoice patterns; add identity/liveness verification steps; ongoing client monitoring and grading. - Operations automation: CRM, lead and onboarding workflows, document handling, and hand-offs between sales, underwriting, funding and collections. - Platform integrations: API connections to our factoring software, CRM, fuel-card and payment data, plus internal credit, portfolio and profitability dashboards. - AI dispatch: an early-stage tool to help small fleets find and plan loads and communicate with brokers. MUST-HAVE SKILLS - 3+ years of Python or TypeScript building production systems - Real-world LLM app experience: OpenAI/Anthropic APIs, RAG, tool/function calling, document extraction from PDFs and scans - Workflow automation (n8n, Make, Zapier or custom code) and REST API integrations - SQL and basic data modeling - Clear written English and short daily/async progress updates NICE TO HAVE - Fintech, lending, factoring or credit-risk experience - Trucking/logistics data: load boards, ELD, FMCSA/DOT - AI agent frameworks or multi-agent orchestration - Fraud detection or ML risk scoring - AWS - A few hours of overlap with Israel time (IST) DEFINITION OF DONE (first milestones) - A working underwriting intake pipeline that extracts key fields from applicant documents with measured accuracy and produces a summary + risk flags for a human reviewer - Documented, tested code in our repo with a short handover note for each feature Engagement: 20-30 hrs/week, 3-6 months, with possible extension.