Responsibilities
- Build multi-agent workflows
- Write system prompts in Italian and English
- Ship to production daily
- Understand the codebase, how agents coordinate, and how workflows reach production
- Pair with Lorenzo on existing workflows — fix bugs, improve prompts, add steps
- Ship your first agent or workflow improvement to production
- Design and build new agents with bilingual system prompts and Zod output schemas
- Wire AI workflows end-to-end: Mastra steps, API routes, React frontend with real-time streaming
- Own prompt performance on your workflows: measure, optimize, iterate
- Architect multi-step pipelines from scratch (parallel execution, concurrency control, error recovery)
- Implement document processing capabilities (PDF, DOCX) with confidence scoring
- Create tools for agents to interact with databases, S3, and external services
- Shape what we build next, not just how
- Have opinions on architecture, push back on ideas that don't make sense, and drive features from 'we should do X' to production
Requirements
- Experience building and shipping AI agents in production
- Proficiency in TypeScript
- Experience with AI frameworks and agent orchestration
- Ability to write system prompts in both Italian and English
- Experience with CI/CD pipelines and deploying to production daily
- Strong understanding of structured output schemas (e.g., Zod)
- Experience with real-time streaming to frontend (React)
- Experience working in a monorepo environment
Nice to Have
- Experience with Mastra Framework
- Familiarity with Azure OpenAI and GPT models
- Experience with Next.js and React
- Experience with PostgreSQL and AWS ECS
- Experience with document processing (PDF, DOCX)
- Experience with bilingual (Italian/English) AI systems
Benefits
- Equity included
- Access to AI tools (e.g., Claude Code subscription or preferred AI tool)
- Flexible working hours with core hours 9:15–18:15 CET
- Fast-paced environment: Slack message > meeting
Work Arrangement
Remote (Worldwide)
Team
Team size: 9. Structure: 9 people total: 4 founders and 5 team members. Engineering team consists of Lorenzo (CTO), a Founding Engineer, and the new hire as the third engineer. Reports directly to both Lorenzo and Daniele (CPO).
The Team
We're 9 people total - 4 founders and 5 team members. Two co-founders work on the product daily: Lorenzo (CTO) and Daniele (CPO). The engineering team is Lorenzo and a Founding Engineer. You'd be the third engineer, working directly with both co-founders.
How We Work
Ship daily - CI/CD goes straight to ECS. Your code hits production the same day. AI-first - We build with Claude Code, Cursor, whatever makes us faster. We'll give anything you need to experiment and build faster (eg. Claude Code subscription or the AI tool you prefer). Pragmatic - Simple solutions, DRY code, iterate based on real usage. No over-engineering. Flexible - Core hours 9:15-18:15 CET. We care about output. Fast - Slack message > meeting. Always.
Tasks Why This Role
Most 'AI Engineer' roles are about wrapping an API call in a chatbot. This isn't that. We have 40+ specialized agents across 5 major workflows already in production - parallel execution, human-in-the-loop conversations, document understanding, real-time streaming to a React frontend. Every agent has a purpose, a structured output schema, and a place in a larger orchestration.
What You'll Do
First month - Learn the system, ship real things: Understand the codebase, how agents coordinate, and how workflows reach production; Pair with Lorenzo on existing workflows - fix bugs, improve prompts, add steps; Ship your first agent or workflow improvement to production. Months 2-3 - Build independently: Design and build new agents with bilingual system prompts and Zod output schemas; Wire AI workflows end-to-end: Mastra steps, API routes, React frontend with real-time streaming; Own prompt performance on your workflows: measure, optimize, iterate. Months 3-6 - Own entire workflows: Architect multi-step pipelines from scratch (parallel execution, concurrency control, error recovery); Implement document processing capabilities (PDF, DOCX) with confidence scoring; Create tools for agents to interact with databases, S3, and external services; Start shaping what we build next, not just how. After that - You're one of three engineers building the product. You'll have opinions on architecture, push back on ideas that don't make sense, and drive features from 'we should do X' to production. The scope grows as fast as you do.
What's Already Built
We're not pre-product. We have a mature codebase with 40+ agents across 5 major workflows: conversational venue search, AI-driven event brief gathering, supplier proposal extraction, and email automation. Real customers use it today. You'll work in a TypeScript monorepo: Next.js web app, Mastra + Hono AI service, shared database and streaming packages, human-in-the-loop patterns with suspend/resume.
Additional Information
- Core hours: 9:15–18:15 CET
- Languages required: Italian and English (for writing system prompts)
- Stack: TypeScript monorepo (Turborepo + Bun), Mastra Framework, Azure OpenAI (GPT-5.2), Next.js 16, React 19, PostgreSQL, AWS ECS
- AI tools provided: Claude Code, Cursor, or any preferred AI tool
