Noxx is hiring on behalf of another AI company We're partnering with an ambitious, early-stage fintech startup on a mission to transform how people manage their money across Latin America. Founded by experienced entrepreneurs with a previous successful exit, the company is building an AI-powered financial agent that combines automation and modern technology to help people make smarter financial decisions and achieve their goals. Rather than expecting people to adapt to complicated financial products, they're creating technology that adapts to people's lives. This is an exciting opportunity to join a small, fast-moving team, work closely with the founders, and help shape an innovative product from the ground up. You'll have real ownership, room to experiment, and the chance to make a meaningful impact on how the next generation experiences financial services.

About the Role
Take full ownership of our agent runtime. You'll build the systems that let a personal finance agent operate safely, reliably, and verifiably in production: memory, context assembly, state management, task execution, tool integration, execution boundaries, evals, and recovery. The runtime is partially built and your first mandate is to understand what's there, cut what doesn't need to exist, and make the rest trustworthy enough to move real money

What you'll do
- Own the full runtime loop: from planner to task execution to verification and recovery.
- Design and implement memory infrastructure: episodic, semantic, temporal state, context assembly, compaction, and conflict resolution.
- Own state machines, session runners, pause/resume, connection state, and challenge detection. Execution integration




- Decide what Pepper owns vs delegates, and build the abstraction layer that makes that contract clear and stable.
- Own credential flows, OTP/auth state, capability permissions, and safe execution boundaries. Reliability
- Build tracing, telemetry, replay, and regression infrastructure so failures leave a trail.
- Implement deterministic replay and dry-run modes: know the blast radius before an action runs.
- Design and run evals: Pepper verifies outcomes.
- Ensure idempotency, retries, and recovery work in production, not just in tests.
- Build human-in-the-loop review workflows for high-stakes actions where the agent must pause and confirm. Working with the Team
- Ownership and agency: shape architecture, engineering standards, product direction, and future technical hires, with room to influence the broader company. Who you are
- High agency and bias toward action: you’re an adventurer, a doer, and a teamplayer. You form opinions, speak up, make decisions, and take a step forward under uncertainty.
- Real agent-building scars: you've shipped agent infrastructure, debugged real failures, and built systems to prevent them from recurring.
- Comfortable entering an existing codebase, forming a view of what's wrong, and becoming its owner without waiting for a roadmap.
- B2C

experience:




you’ve built for consumers and shipped products that real people used. Skills we're looking for
- What you have done and can do > where you studied or worked before.
- Agent runtimes in production: you've built multi-step agents that run reliably at scale.
- Systems thinking: async/distributed systems, retries, idempotency, long-running tasks, state machines, and recovery are native concepts for you.
- Memory and context: you understand persistence, retrieval, temporal state, context assembly, and compaction; and you've built the plumbing, not just used the APIs.
- Evals as discipline: you design evaluation harnesses for non-deterministic systems and think in regressions, not just unit tests.
- Determinism vs LLM judgment: you know when to use each, and you build accordingly. This is not the role for you if…
- Your agentic experience is primarily prompting, RAG pipelines, or wrapping LLM frameworks.
- You've built AI-powered features but haven't owned the runtime or infrastructure underneath them.
- You need a well-defined ticket queue to be productive.
- You'd describe yourself primarily as a QA engineer, frontend engineer, or AI product engineer. You can expect
- A small and ambitious team that will challenge you and push you forward.
- A fast-paced environment to learn and grow continuously; speed over perfection, always
- Founding team: a spirit of high agency and ownership, no middle layers.
- On-site work at our HQ in Del Valle, a few blocks from WTC.
- Build a new category 0 to 1: no playbooks, no hand-holding, no limits, lots of passion, creativity, and obsession

📌 Founding Engineer, Agentic Systems (Colombia)
🏢 Noxx
📍 Colombia