Software Engineer - AI Automation Systems

ABOUT VELEIRO Veleiro AI is the AI delivery and execution system for IT service firms. Our agents work across the full project lifecycle, from discovery through definition, design, and delivery, drafting proposals, statements of work, architecture, and configuration inside our customers' own systems, with a consultant approving every step. We are a fast-growing company based in Silicon Valley, building next-generation agents that automate workflows end to end. WHY JOIN US If you want to build AI agents that do real work for real customers, own what you ship, work with frontier technologies and grow as fast as the company does, we would like to meet you. ABOUT THE ROLE We are looking for a Software Engineer to build the AI agents and the systems that power the AI-driven automation inside our customers' systems. You will design and ship agents that carry a piece of work end to end, integrate them with the enterprise platforms our customers already run on, and build the product surface where consultants interact with our platform. You will also help shape solutions in a fast-moving environment. We need someone who stays up to date with the latest trends, someone who can adapt to any problem and learn whatever technologies are needed to solve it. WHAT YOU WILL DO Agent Development: Design, build, and ship AI agents that complete real tasks end to end, including tool design, prompting, state, and control flow.Reliability and Guardrails: Make agent output trustworthy enough for client-facing delivery. Define guardrails, approval points, and what an agent does when it fails.Evaluation: Build the datasets, regression checks, and traces that turn agent quality into a number you can act on.Platform Integration: Connect agents to the enterprise systems our customers run on, with real attention to authentication, permissions, and data boundaries.Full-Stack Delivery: Own features from data model and API through to the interface where consultants review and approve agent work.Multi-Tenant Security: Enforce data isolation between customers at every layer of the system.AI-Augmented Development: Use AI coding agents as a daily instrument, with the judgment to challenge and override their output when it compromises maintainability.Product Judgment: Work directly with the CEO and CTO to turn loosely defined customer problems into scope you can ship. MUST HAVES Strong programming skills in any object-oriented language.Relational database fundamentals (PostgreSQL or equivalent): schema design, migrations, and query performance.Working knowledge of LLM APIs (Anthropic, OpenAI, Google) and of agent fundamentals: tool calling, structured output, context management, and a real sense of how and why agents fail.Practical, hands-on use of AI-assisted development tools (Claude Code, Cursor, Codex or similar), paired with the critical judgment to reject their output when it is wrong.Attention to detail that survives production: you check the edge case and you read the actual error.Proactive ownership: you identify what needs to happen next and take it.Comfort with ambiguity: work is frequently underspecified. This role requires defining the problem before solving it.First-principles thinking: when the standard path does not fit, you say so before building it, and you are comfortable working around a system that blocks you.Upper-intermediate English level. All code, documentation, specifications, and pull requests are in English. NICE TO HAVES Production experience with LangChain or LangGraph.Python production code, REST API design.Solid TypeScript and React: component design, state management, and comfort owning a feature all the way to the browser.Agent-Driven Development: subagents, skills, spec-first workflows, isolated worktrees, agent-reviewed pull requests.Model Context Protocol (MCP): building servers, clients, or tool schemas.Eval design for LLM systems: golden datasets, LLM-as-judge, regression suites, tracing with LangSmith or similar.Retrieval-augmented generation and vector search (pgvector, LanceDB, Pinecone or similar).Django, Celery, or asynchronous task processing at volume.AWS, containerized deployments, and infrastructure as code.Observability for AI systems: traces, token and cost accounting, rollback.Familiarity with Salesforce or another CRM/ERP platform.Multi-tenant SaaS experience. PERKS AND BENEFITS Growth without a ceiling: we are scaling fast, and scope expands as quickly as you can absorb it. Nobody will hold work back from you.Frontier tooling: you will work with the newest models and agent frameworks the week they ship, because the product depends on it.Real ownership: you own systems and outcomes end to end.A direct line to leadership: you work with the CEO and CTO without layers in between.A bar that raises yours: every specification and pull request gets a serious review before it ships. That review is where most of the growth happens. HOW WE HIRE Transparent by design. Application. A form, plus one unusual ask: send us a raw AI coding session. Unedited. We want to see how you prompt, how you correct course when the model drifts, and what you accept or reject. Build Day. Saturday, August 29, in Medellín. A full paid day building alongside our team on a real project. Lunch included. You compete for prizes. Interview with the CEO. The strongest performers on Build Day go straight to this conversation. Offer. Applications close Sunday, August 23, at midnight Colombia time. We answer every applicant on Monday, August 24, whether you make it through or not. LOCATION This is a hybrid role. The team works remotely most of the time and meets in person occasionally in Medellín for planning sessions, code sprints, Build Days, and team gatherings. Candidates should be based in Medellín or be able to attend these in-person sessions when needed.