Forward-Deployed AI Architect

We are looking for a hands-on Forward-Deployed AI Architect to lead the technical delivery of an AI-assisted operations platform for an enterprise client in travel disruption management.

The current phase covers three connected workflows:

• Processing and routing service requests

• Ranking and recommending hotels

• Closing cases with billing reconciliation and downstream system handoff

You will turn operational knowledge and existing software into a working, testable system with explicit human approvals and traceable decisions.

This is technical leadership, not architecture-only advice or task execution. You will work directly with operational experts, the client’s technology partners, and our engineers, challenge assumptions constructively, and carry the technical solution from discovery through implementation and evaluation.

How you will work

You own the technical approach, engineering plan, implementation quality, and evidence of readiness. You partner with the engagement Principal, who owns the integrated delivery plan, commercial scope, and client commitments. You provide feasibility assessments before commitments are made and escalate unresolved dependencies or scope changes with options and their impact.

What you will own

1. Establish an evidence-based technical baseline

• Inspect the code, infrastructure, APIs, and operator workflows before estimating the build.

• Translate discovery into concise specifications with inputs, outputs, exclusions, acceptance tests, and source references.

• Distinguish confirmed requirements from analyst recommendations and assumptions.

• Resolve contradictions with the appropriate owner. Do not implement them silently.

2. Make integration work early

• Build and review critical components, guide engineers, and prove a thin end-to-end path in the client’s test environment: receive a request, retrieve operational data, produce a recommendation, obtain approval, and perform the agreed system handoff.

• Verify API access, write capabilities, event contracts, and error paths. Do not treat documentation or isolated components as proof of readiness.

3. Make rules and knowledge maintainable

• Design a clear separation between operational facts, business policies, and software logic.

• Work with named business owners to define rule precedence, exceptions, validation, and approval.

• Keep draft or inferred rules separate from approved rules.

• Support versioned updates without a software release, with historical replay, traceable approval, and rollback.

• Use grounded reasoning and human escalation where rules do not resolve a case.

4. Own engineering quality and evaluation

• Review and test code, including AI-generated code, against the agreed architecture.

• Establish automated CI checks, integration tests, and representative operational cases.

• Track the exact application, rule, and data versions evaluated.

• Implement decision logs, monitoring, access controls, and failure recovery.

• Define test-readiness evidence and agree workflow-quality, latency, and operating-cost measures with the team and client.

5. Keep technical decisions and progress clear

• Maintain a concise architecture and decision record that shows what works, what remains, and why.

• Raise blockers early with evidence, an owner, a decision needed, and delivery impact.

• Propose simpler alternatives and update estimates when assumptions change.

• Demonstrate working increments and clearly distinguish prototypes from integrated functionality.

• Leave the team with maintainable code, documentation, and a practical handover.

What good looks like

• Before build commitments: a reviewed technical baseline, verified critical dependencies, and explicit assumptions behind the estimate.

• During delivery: a working end-to-end increment, automated tests running in CI, and demonstrated progress against agreed acceptance cases.

• For rule changes: a named business approver, a versioned baseline, and recorded replay results before promotion.

• At evaluation: reproducible results, decision traces, and documented limitations that support the agreed Go/No-Go decision, not just a successful demo.

Specific targets will be agreed after the baseline is established. You will stay involved from the initial working sessions through delivery and evaluation.

What you bring

• Evidence of taking an applied AI or LLM-enabled workflow beyond a prototype into operational use.

• Strong hands-on software engineering, architecture, API integration, and code-review skills.

• Experience turning incomplete or conflicting discovery material into implementable, testable specifications.

• Sound judgment about structured rules, retrieval, AI reasoning, and human review, including their failure modes.

• Practical experience with evaluation datasets, regression testing, CI, and production observability.

• Confidence to challenge assumptions, explain alternatives, and escalate decisions constructively across business and engineering teams.

• Clear, concise communication and realistic estimation based on verified dependencies and available capacity.

Nice to have

• Experience in exception-heavy operations, business-rule systems, travel, hospitality, or logistics.

• Familiarity with AWS-based AI deployments.

Logistics

The role includes an in-person working session in Milan. Please share your earliest start date, realistic weekly availability, and ability to travel.