*Native/Bilingual English is required for this role (read/written/spoken)Please upload your CV Resume in English. Monthly salary:$5,000 - $6,500 USDAlong with our partner, we are seeking an engineer who takes prototypes to production: hardening, deploying, iterating with real users, and owning the full delivery pipeline. You're the person who makes AI products real.The role: They have prototypes that work. What they don't have is a dedicated engineer to take them to production. You'll take existing foundations, harden them into production-grade applications, deploy them, and iterate based on real user feedback. You'll also prototype new AI-powered features and capabilities alongside your director.This is an engineering execution role, not a product management role. You'll make technical decisions about AI behavior, implement LLM tool-use flows for non-technical hotel operators, and own the entire delivery pipeline from working prototype to deployed, monitored, production application.What you'll do: Take prototypes to production.The prototypes exist a conversational BI tool with NL2SQL and a portfolio-level operations dashboard. You'll add auth, monitoring, error handling, test coverage, and deploy them for real users. Own AI engineering decisions.Implement how the LLM interacts with users tool-use flows, NL2SQL generation, streaming UX, error recovery, conversation lifecycle. You have opinions about this, not just implementation skills. Build full-stack.Angular 19/21, Fastify 5, TypeScript end-to-end. Frontend components, API routes, Snowflake queries you move across the stack as needed. Harden and deploy.CI/CD pipelines (Azure DevOps), Azure AD authentication, error handling, test coverage, performance optimization. Make prototypes production-safe. Close the loop with stakeholders.Work with their director and the product team to translate user feedback into engineering iterations. You won't be customer-facing, but you'll be close to the feedback and responsible for acting on it. Manage and extend the verified query set.Their NL2SQL accuracy is powered by example-led generation a curated library of verified queries that constrain LLM output. You'll add to it, run gap analyses, and use it as the primary lever for improving answer accuracy. Close the feedback loop.User signals acknowledgments, ratings, corrections feed back into system quality. You'll design pipelines that route those signals to verified query sets, prompt refinement, or evaluation frameworks so the system improves measurably with use. Work with AI agents.Their development workflow uses Claude Code with specialist AI agents, CLAUDE.Md-driven conventions, and Design → Build → Review pipelines. You'll adopt and improve this system.Tech you'll work with: Application layer:TypeScript, Angular 19/21, Fastify 5, Node.Js, React Monorepo and standalone app structures. SSE streaming, Zustand/NgRx state management, ECharts, Material UI. AI / LLM integration:Gemini (Vertex AI), Claude API, Tool-use patterns, SSE streaming, MCP (Model Context Protocol) Text-to-Specification patterns the LLM classifies and parameterizes against a semantic layer rather than writing raw SQL. dbt semantic layer for metric definitions and dimension catalogs. MCP server architecture serving data retrieval across multiple products. Data layer:Snowflake, dbt, Data Vault 2.0 You don't need to be a data engineer, but you'll write Snowflake queries and understand the Data Vault model backing their analytics. Infrastructure:Azure, Azure DevOps, Azure AD CI/CD pipelines, cloud hosting, SSO authentication. You'll own the path from merged PR to running in production.What they're looking for: Must-haves AI engineering experience. You've built and shipped applications that use LLMs not just wrappers around ChatGPT, but real tool-use flows, structured generation, or agent systems. You have opinions about when to stream vs. batch, when to ask for clarification vs. infer, and how to handle LLM errors gracefully. Full-stack TypeScript. Strong in Angular (their primary customer-facing framework). React experience is a plus but Angular is the requirement. Solid Node.Js/Fastify backend skills. You can own a feature from database query to rendered UI. Production shipping muscle. You've taken something from "it works on my machine" to "it's running in production with real users." CI/CD, monitoring, auth, error handling the foundational work that makes software real. Direct communication. You ask questions before building. You push back on vague requirements. You say "this seems wrong" when it does. No yes-people. AI agent systems experience. You've built, managed, or worked extensively with AI agent systems multi-agent orchestration, tool-use pipelines, or agentic development workflows. Their entire dev process runs through 30+ specialist AI agents; you need to be comfortable operating in and improving that system. AI-assisted development. Comfortable working with AI coding tools (Claude Code, Copilot, Cursor, etc.) as a core part of your workflow, not a novelty. Multi-tenant AI safety. You understand the difference between enforcing data isolation at the LLM layer (fragile) versus the data connection layer (robust) and why that distinction matters when customer data can never cross boundaries.Nice-to-haves: Hospitality industry experience or multi-property portfolio analytics Snowflake / SQL analytics experience Experience with SSE or WebSocket-based real-time UX Azure cloud platform experience LLM observability and evaluation frameworks (LangSmith, Arize Phoenix, or similar) instrumenting AI pipelines to track faithfulness and relevancy, not just uptimeHow our partner works: Their development workflow is AI-native. They use Claude Code with 30+ specialist AI agents, CLAUDE.Md project conventions, and a Design → Build → Review pipeline. Every feature goes through product design agents before coding and security + QA agents after. This isn't optional it's how they ship quality at speed with a small team. Small team, high autonomy. You report directly to the Director of Data Engineering. Minimal process, maximum ownership. Real data only. No mock data in any environment. APIs return errors when Snowflake is down, not fake arrays. Empty states say "not yet available," not fake numbers. Documentation as code. CLAUDE.Md files, architecture docs, and API route docs are living documents that stay in sync with the codebase. Quality gates. Security review + QA review after every implementation round. Non-negotiable.Benefits: A fully remote position with a structured schedule that supports work-life balance. The opportunity to build production-grade AI products at the forefront of agent-driven technology. Two weeks of paid vacation per year. 10 paid days for local holidays.*Please note our partner is only looking for full-time dedicated team members who are eager to fully integrate within their team.