Luca Lillo · Chief Technology Officer
Psychology-Informed Governance | Evidence-Led Evaluation
I design and govern enterprise AI systems where decisions must be clear, testable, and accountable. Currently, I serve as Chief Technology Officer, combining technical leadership, engineering governance, and team management with a focus on building, governing, and evaluating LLM multi-agent systems.
My work explores how agent structure, role design, authority, information flow, structured dissent, and human oversight impact decision quality. My background includes 18 years at Enel Group across enterprise technology, electricity distribution, digital transformation, and programme delivery. My dedicated focus on AI began in 2025.
My academic training in Psychological Sciences gives me an engineering perspective on authority bias, conformity, groupthink, social loafing, and overconfidence. I do not treat language models as humans, nor do I assume they have human intentions. Instead, I translate these cognitive dynamics into concrete failure risks, structural safeguards, and testing strategies.
My core research areas include separating procedural control from reasoning authority, embedding structured dissent and falsifiable objections, running blind reviews and external checks, testing agent ablations, and comparing multi-agent setups against single-agent baselines with equal compute. I measure performance based on accuracy, resistance to misleading prompts, error propagation, traceability, and cost.
Scope: These are ongoing conceptual and experimental designs, not a validated industry framework or completed benchmark.
Core Question → Literature Review → Stated Assumptions → Success & Failure Criteria → Instrumentation → Validation.
I review current academic and technical research, clearly separate proven facts from working hypotheses, and set clear evaluation criteria before claiming results. I instrument systems to monitor quality, cost, latency, and failure rates, refining architectural choices based on real data and known limitations.
I authored Sistemi & Automazione’s internal guidelines for AI-assisted software engineering. These guidelines turn academic findings into practical controls for requirements, architecture, code implementation, testing, traceability, and human accountability.
Boundary: These are internal company guidelines, not a public standard or an externally audited framework.
Sistemi & Automazione S.p.A. · June 2026–present · Rome, Italy
I lead technical strategy, engineering standards, project governance, technology selection, and software delivery quality. My remit covers organisational design, vendor qualification, and technical reviews for major proposals and contracts. I directly manage 5 team leads and guide 15 additional contributors across functional lines.
I bring 18 years of experience across enterprise systems, electricity grid automation, digital transformation, large-scale programme delivery, and team capability development.
Experience boundary: My dedicated technical specialisation in AI began in 2025.
A controlled agent workflow demonstrating sovereignty-aware model routing, human sign-off, full decision lineage, and runtime telemetry.
Limitation: A structured, typed reference workflow with governed model calls—not a general-purpose autonomous runtime or multi-tenant production engine.
View Decision Matrix repository (opens in a new tab) · Open live demonstration (opens in a new tab)
A policy-governed foundation for AI-assisted development using schema contracts, explicit state transitions, PostgreSQL leasing, and fail-closed security controls.
Limitation: An open architecture and core PoC, not a complete turn-key production platform.
A portfolio project featuring temporal validation, reproducible training pipelines, MLflow artefact tracking, Docker packaging, and automated GitHub workflows.
Limitation: The accompanying live FastAPI endpoint is rule-based. It is an independent portfolio demonstration, not an Enel production system.
Neuromorphic Inference Lab is my independent open-source research initiative. I define system architectures and specifications, oversee AI-assisted implementation, review all code and tests, and maintain full accountability for validation, limitations, and deployment choices.
My public portfolio showcases inspectable architectural patterns, bounded PoCs, and experimental evaluation designs. I do not represent this work as enterprise client deployments, autonomous multi-agent engines, or peer-validated industry benchmarks.
Italian/EU citizen · No German visa sponsorship required · Open to relocation across Germany upon an accepted offer
