AI Engineer | AWS Bedrock + React/Node.js + Elasticsearch | Remote

Role: AI Engineer Duration: Long Tern Location: (Remote) Job Description Black Book is looking for a contract software engineer to lead the build-out of the AI assistant inside Pulse. Pulse pulls in automotive vehicle listing data from across the US and Canada, summarizes up by VIN, and turns it into the analytics our customers use day to day. Our next step is making all that available through natural language: an assistant (agentChat, built on AWS Bedrock) that lets customers ask questions about their inventory and market and get real answers back, safely and within their existing permissions. This is the core of the role. We want someone who will own the assistant end to end, from prompt and agent design through the tools and data it can reach, the guardrails around it, and the way it surfaces in the product. There is plenty of supporting full-stack work around it (React front end, GraphQL/Node services, Elasticsearch, AWS), but the assistant is the priority, not a side project. Just as important is how you work. We build with AI. Our engineers use Claude Code and an AI-DLC-style, spec-driven workflow to plan, generate, and review code, and we expect you to do the same. We are not looking to add someone to hand-write code line by line. We want an experienced engineer who can direct AI coding agents well: break work into clear specs, steer the tools, and review and validate what comes back with a critical eye. You still need to deeply understand the stack, because you must judge whether the AI got it right. This isn’t a junior role. We need someone who has already built and run full-stack applications on AWS, has shipped with LLMs, and is comfortable working in a fast-paced environment. What You’ll Be Doing Expanding the Pulse AI assistant (agentChat on AWS Bedrock): agent and prompt design, the tools and data it can call, and the guardrails that keep answers accurate and inside each customer’s permissions. Turning fuzzy product questions into a reliable assistant: deciding what it should be able to answer, wiring it up to Elasticsearch analytics and the rest of the platform, and steadily improving quality. Evaluating and improving the assistant (accuracy, safety, latency, cost) and putting evals and feedback loops in place so we can measure it. Working in an AI-native way: using Claude Code and an AI-DLC-style, spec-driven workflow to plan, generate, and review code rather than writing it by hand. Reviewing and validating AI-generated code carefully, testing it, and taking responsibility for what ships. Building the supporting full-stack work around the assistant across Pulse: React 19 front end (Apollo Client, styled-components, NextUI) and the Apollo GraphQL server on Express and Node 22. Working with the analytics and data layers the assistant relies on: Elasticsearch queries and aggregations, the SQS/SNS/Kinesis pipeline and background workers, and data stores including MS SQL Server (via Sequelize), DynamoDB, and Redis. Handling JWT authentication and the company- and country-level access rules that keep tenants separated, since the assistant must respect all of it. Working the way the team does: Gitflow branching, code reviews. REQUIRED SKILLS/EXPERIENCE: At least 6 years building and running enterprise web applications. Hands-on experience building with LLMs and agent-style features (AWS Bedrock, or similar), including prompt and tool design and putting guardrails and evaluations around them. Comfortable working AI-native: real experience using Claude Code (or a similar agentic coding tool) as your main way of building software, plus an AI-DLC or spec-driven workflow. You direct the tools and review the output; you don’t hand-write everything. The judgment to review and validate AI-generated code and catch what’s wrong, which means genuinely understanding the stack underneath it. Strong TypeScript and solid hands-on React experience (hooks, single-page apps; we’re on React 19). Good backend experience with Node.js and GraphQL (Apollo), and comfort with schema-first design and code generation. Real AWS experience across a range of services, including ECS, S3/CloudFront, SQS, SNS, Kinesis, not just the application code that sits on top of them. Hands-on Elasticsearch experience with queries, aggregations, and index management; this is a big part of how the product works and what the assistant draws on. Experience with relational databases and ORMs (MS SQL Server and Sequelize) as well as DynamoDB and Redis. Experience designing and working with event-driven, queue-based systems at scale. A genuine testing habit with Jest (and tools like React Testing Library and supertest), and no complaints about linters, formatters, and strict type checking. Comfortable working in a monorepo (Lerna and npm workspaces) and with Docker for local development. Able to work independently, own features start to finish and stay effective in a codebase that has some legacy and tech debt. Clear written and verbal communication; the team is remote and distributed. Nice to Have Experience with LLM evaluation, RAG, or agent orchestration frameworks. Mapbox GL and geospatial work (@turf/turf, GL Draw). Charting and data-viz libraries like Nivo or TanStack Table. OpenTelemetry, or Storybook. Experience with multi-tenant B2B SaaS and internationalization (i18next). Familiarity with the automotive or vehicle-data space. Comfortable working in Windows and PowerShell, which is what most of the team uses. All your information will be kept confidential according to EEO guidelines.