Head of AI Engineering (Architecture & Agents)

Help 350,000 travellers a year move across Europe

Location: Remote (CET) with two days/month on-site or Ljubljana, Slovenia · Full-time

Reports to: CEO, working daily with the dev team and CTO · Start February 2027

Salary: €75,000 - 100,000 gross per year + bonus and equity (TBC)

GoOpti is a door-to-door shared-transfer marketplace connecting cities and airports across Europe. Behind every booking sits real-time search, dynamic pricing, vehicle occupancy optimisation, franchise and partner settlement, and customer care in several languages and countries.

We have decided to become an AI-native company. Not a chatbot bolted onto the website - a business where agents do real work in every department, safely, measurably, and connected to the systems that actually run the company.

We are looking for the person who designs that foundation and ships the first agents on top of it. You will be our first senior AI hire, with a direct line to the CEO and CTO and an open mandate to shape how we do this.

The role

You will do two things at once:

1. Architect the agent platform - how agents are orchestrated, how they reach our internal systems and data, how they handle context and memory, how we monitor them and where a person stays in the loop. An architecture that makes the tenth agent a week of work, not a new project.

2. Build and ship - hands on the code, connecting AI to our Java backend, our APIs and our databases, and getting the first autonomous agents into daily use.

What you'll own

The Company Brain

Bring together bookings, prices, trips and franchise results from our data warehouse with our business rules (fares, refunds, franchise terms) and our working knowledge - MS Teams, email, customer-care conversations, complaints, Git - into one machine-readable knowledge layer that agents can query, that flags contradictions and keeps itself up to date.

Agent platform architecture

• Tool / MCP layer over our booking, pricing, planning, analytics and financesystems
• Orchestration and hand-off between agents, short- and long-term memory, context management
• Observability, tracing and evaluations for non-deterministic systems
• Human-in-the-loop checkpoints and permissioning — a person approves anything touching price or customers

The autonomous agents - start where impact is highest, prove it, expand:

• Marketing - content, SEO/AEO and AI-readable answer pages across our markets
• Development - agents supporting our engineering workflow, from triage to merge requests
• Finance - agents supporting cashflow forecasting, reporting and clearing/refund reconciliation
• Customer care - booking changes handled end to end by our customer-care chatbots (live today, built on OpenAI)
• Pricing & planning - assistants that capture and apply the knowledge of our pricing and planning experts
• Sales (fast follow) - lead handling and B2B partner support

Self-improving loops & token economics

When an agent fails, a supervisor agent diagnoses why, drafts the code fix or policy change, tests it in a sandbox and opens a merge request for a person to approve. Own the model and token budget, choose models per task, run evaluations, and report cost per booking.

Intake, priorities & people

Run a weekly intake of problems from department heads — at most two builds per department at once,

each with an owner, a measure and a go/no-go gate. Lead one or two contract AI developers, work alongside our backend team, and teach managers how to write a good spec for an agent.

What success looks like

• By 31 March 2027 - the platform foundation in place; three agent skills in daily use; baselines set; a supervisor agent learning from every conversation our chatbots hand to a person; our pricing and planning knowledge captured in the Brain.
• By 30 June 2027 - at least two skills live in every department; improvement loops running in customer care, complaints and planning.
• By 30 September 2027 - the summer peak handled with office headcount flat; an AI-native review of the whole company.

How we'll measure it: hours freed per week · share of customer conversations resolved without a person · revenue per employee · token cost per booking · fixes merged from loops. Targets are set together from the baselines.

What you'll connect AI to

• Backend: Java core marketplace and micro services; new services in **Java** (REST, Hibernate/Panache, REST clients); legacy JRuby/Rails storefront being migrated to React
• APIs: internal REST services, partner and aggregator APIs, payment and fraud screening integrations
• Data: MySQL (operational + analytics databases), PostgreSQL, Redis, DynamoDB clickstream, a new BI warehouse
• Infrastructure: AWS (EKS), Docker, Kubernetes / Helm, GitHub Actions CI/CD, New Relic

What we're looking for

• 6+ years building backend or data systems, including 2+ years shipping LLM-based systems in production - tool use, retrieval, evaluations, agents. You have hit context limits, tool-calling failures, runaway costs and silent regressions, and you have opinions about each.
• Strong backend engineering - you are comfortable in a large Java codebase, design clean REST APIs and integrate with existing services. Python or TypeScript for AI tooling.
• Solid data modelling - SQL you can reason about seriously (MySQL / PostgreSQL)
• Architecture thinking - you can argue for and against an approach, write the decision down, and design for scale, security and GDPR from day one.
• Pragmatism - a record of building practical things fast in a small company with real, existing systems.
• Communication - you explain trade-offs to non-technical managers and know when to say no. Fluent English.

Nice to have

• Building MCP servers and multi-agent orchestration in production
• Experience with LLM observability / eval tooling and model routing across providers (OpenAI, Anthropic, open models)
• Travel, mobility or marketplace experience; routing or operations research