About Travezy
Travezy is an AI-powered travel planning product. We run a consumer app and are building a B2B intelligence layer for travel agents. We're small, we ship continuously, and we work with an external engineering partner on core build. We're hiring our first dedicated AI engineer to own how AI actually gets built, evaluated and deployed here.
The Role
Most companies are still at the demo stage with AI. We're past that and hitting the harder problem: getting agentic systems from pilot into production and keeping them reliable there.
You'll own that path. Concretely:
Agentic development and infrastructure
- Build production systems using agentic coding tools (Claude Code, Codex, Cursor, Devin) as a primary workflow, not an experiment
- Design and run the CI/CD pipeline for AI-generated code — how it gets reviewed, tested, merged and deployed safely
- Build the agentic infrastructure our features run on: orchestration, tool access, failure handling, observability
Evaluation and governance
- Build the evaluation framework that tells us whether an AI change made things better or worse, for both generated code and product output
- Own the governance layer for our multi-model setup (we route across several LLM providers) — cost, latency, quality, fallback, and which model is trusted with what
Data and knowledge layer
- Build the semantic layer over our product data. We currently pull from 15+ sources including Mixpanel, GA4, RevenueCat and AppsFlyer, and they don't speak to each other
- Design the ontology that makes that data queryable in natural language
- Stand up MCP servers so our internal tools and agents can reach this data safely
What we are looking for
- You have taken an AI system from pilot to production and kept it running. Not a demo, not a hackathon — something real users depended on.
- You write code with agentic tools daily and have opinions about where they fail
- You've built or meaningfully worked on evaluation for an AI system
- You can reason about data modelling and semantics, not just pipelines
Strongly preferred:
- Experience with MCP, agent orchestration frameworks, or equivalent
- Multi-model routing or LLM gateway experience
- Ontology or knowledge-graph work
- Infrastructure and deployment ownership (we're on [AWS — confirm])
We're not screening on degree, institution, or years.
Process
- Send us your application (below)
- 45-minute portfolio call — you walk us through something you built
- Paid evaluation task on a real problem of ours
- Conversation with the founding team, offer