Full-Stack AI Engineer: LLM Agents, AWS Deployments, Financial Data Integrations

Description: We're an AI-native finance company that builds a strategic-finance platform and deploys it into our clients' cloud environments, then runs it as a managed service. Small senior team, fast ship cycles, heavy use of agentic tooling (Claude Code) end to end. We're adding a contract engineer who can own builds from architecture to deployed-and-working. What you'll build - Packaged, repeatable deployments of our platform into client AWS accounts — containers, S3, IAM/credential handling done cleanly - Data connectors and pipelines for finance systems (QuickBooks Online, banking/payments APIs, time-tracking APIs) and external sources (market/fundraising data APIs) - Extensions to our modular app/demo builder: dashboards, scenario and reporting views over structured finance data - Agentic LLM workflows — deterministic flows where accuracy is load-bearing, eval layers where it matters - Hardening: secrets, isolation, audit-friendly practices (we're on a path toward SOC 2) - Reusable modules — every client build should make the next one faster Engagement Independent contractor (1099) via Upwork. Hourly, [$50–$90/hr — set your bracket] depending on experience and fit, ~20–30 hrs/week to start with a long runway for the right person, plus discretionary success fees tied to engagement outcomes. You set your own hours — we align on weekly targets and demo checkpoints, async-first (Slack) with a couple of short team calls each week. Many of our team run their own practices; other clients are welcome. Must-haves - LLM/agent systems you built and shipped to production — and can walk us through: what it did, what broke, how you fixed it - Strong AWS: containerized deploys into accounts you don't own, S3, IAM, networking basics - Full-stack fluency: Python and/or TypeScript on the back end; clean modern front ends - Deep API-integration experience: OAuth, webhooks, rate limits, unreliable third-party data - You can use AI tools heavily — we do — but you must be able to explain and defend every line you ship. Reviews here start with your TLDR: what you built, why, and exactly how the connectors work. - Strong written English Nice-to-haves: finance/accounting domain exposure (month-end close, FP&A, the QBO ecosystem) · Supabase/Postgres · MCP servers or Claude Code power use · prior work in SOC 2 environments · US-overlapping timezone To apply: keep it short — (1) one production LLM/agent system: your role, what it did, what failed; (2) one AWS deployment you owned end to end; (3) rate + weekly availability. Start your proposal with the word LEDGER so we know you read this.