When should a cache change policy?
A warm-up selector evaluated across public traces and synthetic traffic overlays. Recomputed results, fixed-policy baselines and the limits of transfer.
Reproducible artifact ↗Gennaro Francesco Landi
Systems · ML infrastructure · Backend engineering
I build systems I can measure and explain.
I’m an MSc Computer Science student at ETH Zürich, majoring in Data Management Systems with a minor in Machine Learning. My work spans cache evaluation, browser-agent measurement and typed AI workflows.
I work as a Core AI Engineering Intern at Chain IQ and as an independent engineering consultant on search and ranking. I hold a BSc in Computer Engineering from the University of Salerno, where I graduated cum laude, and completed concurrent studies in Statistics for Big Data.
Available full-time from summer 2027, with flexibility on start date and duration.
Selected research & engineering
BrowseTrace
Request-level traces of browser-mediated AI traffic. An open dataset and a cache replay that separates scripted controls from LLM-driven workloads.
Code & data ↗ClaudeFlow
Typed AI pipelines with loops, branching and maps. Zod schemas, execution traces and a deterministic mock runtime for testing orchestration.
Source & benchmarks ↗AuditChain
A team-built procurement prototype that combines language-model parsing with deterministic decision rules and an auditable execution path.
Public hackathon submission ↗Explore the results and project pages →
Earlier projects
V2X co-pilot data platform: traffic and weather adapters, TTL caching and fuzzy risk scoring for my bachelor’s thesis. TORCS: behavioral cloning with nearest-neighbor search. Broletter: an adaptive science newsletter.