# Lorenzo Germano

> Industry & Enterprise Analyst at Accenture's Reinvention Partners practice in Milan, Italy. Background in mathematics, data science, and computer vision; currently focused on agentic AI architectures, multimodal systems, and end-to-end AI solution design for enterprise clients.

This file describes the content of [lorenzogermano.com](https://lorenzogermano.com), Lorenzo Germano's personal portfolio site. The site itself is a single-page React application; this file mirrors its content in plain text for AI/LLM consumption.

## About

- **Name:** Lorenzo Germano
- **Role:** Industry & Enterprise Analyst @ Accenture — Reinvention Partners (current)
- **Location:** Milan, Italy
- **Email:** lorenzogermano1@outlook.it
- **LinkedIn:** https://www.linkedin.com/in/lorenzogermano/
- **GitHub:** https://github.com/jwings1
- **Portfolio:** https://lorenzogermano.com

## Experience

### Industry & Enterprise Analyst — Accenture, Reinvention Partners (Milan, IT — Present)
- Delivered an AI-driven operational-planning solution automating terrain analysis and route recommendation, extended with 2D/3D scenario modeling and real-time situational awareness from geolocated video.
- Architected an agentic multimodal platform that ingests audio and text to auto-generate structured reports, presentations, spreadsheets, and emails via an LLM orchestrating search, database, and automation modules within a secured execution environment.
- Delivered client-facing technical presentations and demos of 5+ POCs leveraging reasoning models, multimodal capabilities, and agent-based architectures across several Public Service clients.
- Led a team of 12 analysts to first place in a practice-wide hackathon (12 teams total) by designing an agentic architecture orchestrating 10+ internal tools.
- Mentored 2 interns in building a customized reality engine using foundation GenAI models for real-time monitoring; deployed to 3 pilot client facilities, cutting alert latency by 70%.
- Ran enablement sessions on SOTA GenAI tools across 2 teams.

## Featured Projects

- [Reinventing national service delivery (Accenture × Poste Italiane case study)](https://www.accenture.com/us-en/case-studies/data-ai/poste-italianes-pivot-postal-service-platform-powerhouse): Enterprise digital reinvention of Italy's national postal service into a multi-service platform serving 45M citizens. Stack: Microsoft Azure, SAP S/4HANA, data mesh, AI agents, computer vision, GenAI tooling. Outcomes include the SuperApp (3.3B annual interactions, #1 Italian finance app), Poste Energia (third-largest energy provider with 1M customers in 3 years), and the Polis Project (~7,000 post offices converted to digital hubs).
- [Pupil Detection from Event Frame Sequence (2025)](https://gamma.app/docs/Pupil-detection-from-event-frame-sequence-zadtehwyksd4aea?mode=doc): Event-based computer vision approach for pupil detection using event camera data, leveraging temporal information and intensity changes for robust eye tracking.
- [P2ROS: Prompt to Robotic Operating System (2025)](https://p2ros-jwings1s-projects.vercel.app/): React-based demo for an AI-powered platform concept that trains robotic arms via imitation learning. Combines consumer-grade 3D scanning (iPhone/iPad LiDAR), computer vision for motion extraction, and AI-powered code generation from natural-language task descriptions.
- Demo XR Tarantismo (2024): Research prototype for immersive cultural-heritage preservation using extended reality, designed for Meta Quest 3/Pro. Models how visitors could experience the southern Italian tarantism ritual in the Chapel of San Paolo delle Tarantate (Galatina) via spatial computing.
- [Palazzo Elefante della Torre (2024)](https://palazzoelefantedellatorre.com/): Custom Lodgify-based website for a luxury vacation rental in an 18th-century baroque palace in Galatina, Italy.
- [3D Human-Object Interaction in Video — H2O-CA (2023)](https://github.com/jwings1/H2O): Research framework for 6-DoF object tracking in monocular RGB video using cross-attention to correlate human pose with object positioning. Implementation in PyTorch. [Project page](https://jwings1.github.io/H2O-CA/).
- [HeimGuard Safety AI Think Tank (2021)](https://jwings1.github.io/HeimGuard/): Conceptualization of an AI-based home safety system combining computer vision, anomaly detection, and IoT sensors for preventive risk identification (falls, cuts, burns, animal incidents, accidental poisonings) — anticipating commercial systems like Aivilon and Sparknets Neurospot.
- [A Simple Digital Filter (2020)](https://github.com/jwings1/A-Simple-Digital-Filter): VHDL implementation of a digital filter using power-of-two coefficients (shift-register datapath, Moore-state-machine controller, overflow detection) for resource-constrained embedded systems.

## Education

- **Claude Certified Architect: Foundations** — Anthropic (Present). Agentic loops, multi-agent systems, MCP server integration, CLAUDE.md hierarchies, structured output schemas, escalation patterns, confidence calibration.
- **M.Sc. Mathematics in Data Science** — Technical University of Munich (March 2021 — March 2024). Focus: probability, optimization, linear algebra; ML for computer vision (multi-object tracking, MOT16), recommender systems, robustness certification, NLP. Implemented Re-ID networks, Message Passing Networks on bipartite graphs, normalizing flows, randomized smoothing, Skip-Gram, spectral clustering, Graph Neural Networks (Cora ML, 85% accuracy), PyTorch image segmentation, RNN/LSTM sentiment analysis, MLOps on GCP. Distributed systems work in C++ (TPC-H query optimization, KNN on directed graphs) and Apache Spark (NYC taxi data).
- **Guest Student** — Denmark Technical University (August 2022 — January 2023). Courses: PDEs, Image Analysis, MLOps.
- **B.Sc. Physical Engineering** — Politecnico di Torino (October 2017 — September 2020). Quantum and statistical mechanics, complex-systems physics, solid-state and nuclear physics; EM-radiation interaction with matter, surface plasmons, photonic crystals; advanced optical lithography (CVD, ALD, etching); intro to quantum computing algorithms (OpenQASM).

## Competitions & Honors

- Anthropic Model Safety Bug Bounty Program (May 2025).
- AI S&Cale Up for Analyst Hackathon, 1st Place — Accenture Italia (March 2025). Led 12-analyst team against 11 competing teams in the "Create Your Agents with Amethyst Studio" challenge.
- Football: Can you guess the winner? Challenge — Qube Research & Technologies (February 2024). Score: 48.86 / 49.32.
- Analytics Cup (DS Competition) — Siemens Advanta Consulting (March 2021). Score: 65.05 / 68.38.
- Athletics National Championship — FIDAL (September 2014). 8th place, Italian Decathlon championship (Allievi category).
- Erasmus+ Grant — TU Munich Department of Mathematics (2022).
- Merit-based Scholarship — Politecnico di Torino (2018–2020).
- National Register of Excellence — Italian Ministry of Education (2017). Perfect score of 100 with honors in the state exam.

## Conferences & Seminars

- 3D Human-Object Interaction in Video — VANDAL Lab, Politecnico di Torino (June 2024).
- Ethics of Responsibility: An Introduction to Applied Ethics — TUM School of Social Sciences and Technology (September 2022).
- Randomized Algorithms — TUM School of Computation, Information and Technology (April 2021).

## Focus Areas

Agentic AI architectures · Multi-agent systems · Multimodal AI (audio, text, vision) · Computer vision · Reasoning and tool-using LLMs · Operational AI for the public sector · Enterprise AI solution design · MLOps · Research prototyping (XR, event-based vision, 3D tracking).