Riccardo
Ceccarani

Natural-language interface for cognitive-decline detection

MSc thesis · EveryWare Lab · in progress

• A multi-agent orchestration layer in LangGraph that lets clinicians question smart-home sensor data in plain language, with state held across separate health domains.
• LLM-generated analysis code runs inside a containerised sandbox — injection risk contained instead of trusted.
• Explainable-AI layer so every answer carries the metric it came from. Non-negotiable when the output touches a diagnosis.

Credit-card fraud detection on a transaction graph

Neo4j · 4M transactions

• Fraud as a graph problem: shared devices, shared merchants, rings of accounts that only look normal one row at a time.
• Cypher tuned with APOC and batch processing until complex pattern matches answered in under two seconds across four million transactions.

Distributed energy supply management

Java · Spring Boot · gRPC · MQTT

• A simulated network of thermal plants electing a leader by Chang–Roberts, with locks and semaphores written from scratch rather than borrowed.
• Multi-threaded by design: concurrent gRPC server threads, MQTT subscribers and sliding-window sensor processing sharing one consistent view.

Snap — short-link analytics for Shopify

Personal project · July 2025

• Full-stack Remix and Prisma app tracking a customer from a trackable link through add-to-cart and checkout, with Shopify Web Pixels feeding real-time events.
• Unique link generation and bot detection, so the revenue a merchant sees is revenue that happened.

Process-mining clustering for large event logs

BSc thesis · Università di Camerino

• N-gram trace representation plus K-Means over event logs, and a repetition-aware distance metric that reads loops correctly where Jaccard flattens them.
• Owned the distance-matrix module inside a three-person framework; parameters validated against the Silhouette index.

Software Engineer

Deloitte

• Backend development with focus on RESTful architectures, microservices, and API design.

EveryWare Lab, Milano

MSc thesis intern · CS research laboratory

• Leading the natural-language interface for SERENADE, an EU-funded project on early detection of cognitive decline from smart-home sensors.
• Architecting the multi-agent core in LangGraph, with state that survives moving between health domains and sensor streams.
• Translating clinical requirements from researchers and clinicians into technical AI and XAI specifications.

Università degli Studi di Camerino

BSc thesis research intern

• Built a process-mining clustering methodology for large-scale event logs and owned the distance-matrix module inside a three-person framework.
• Acted as the link between the algorithmic design and the team's integration work, keeping modules compatible and on schedule.

Milano & Camerino

MSc in progress · BSc 103

• MSc Computer Science (LM-18), Università degli Studi di Milano — distributed and pervasive systems, parallel and distributed algorithms, next-generation data models and DBMSs, artificial intelligence.
• BSc Informatica (L-31), Università degli Studi di Camerino — algorithms and data structures, software engineering, database systems, operating systems, probability and statistics.