2025 Generative AI Projects

1. Email Classifier for Back Office (In Production)

Extended Description

An intelligent email routing system designed to automatically sort incoming messages received by shared back-office mailboxes across the country. These emails cover a wide range of topics, including reimbursements, bonuses, school expenses, contracts, and promotions. The solution combines a customizable rules engine for transparent filtering and a generative AI (Gemini) module to classify more ambiguous or complex cases. The system significantly reduces manual workload and processing time while retaining full control over categorization logic.

Key Components

• Customizable rule engine for initial routing

• Generative AI classification engine for complex cases

• Automated validation and continuous error logging

KPIs and Expected Impact

• ✅ 99.57% classification accuracy across 1,247+ emails, with only a 0.43% error rate (vs. a 5% target)

• ⏱ ~80% reduction in manual email handling time

• 💸 Estimated 70% cost reduction vs. legacy API-based solutions

• 🔍 Improved traceability and accountability in back-office processing

2. AI for Legal (MVP)

Extended Description

An AI-powered solution supporting legal teams by automating the review and summarization of weekly legal updates from official sources such as the Gazzetta Ufficiale and Eur-Lex. The system integrates a generative model (Gemini) to produce relevance-based ratings and summaries, while offering an advanced interactive interface. Users can freely query the AI, perform semantic search across legal documents, compare regulations side-by-side, and provide feedback on accuracy and relevance — enabling a fully active and exploratory legal research experience.

Key Components

• Interactive LLM-powered legal research notebook

• Semantic search and document comparison across official sources

• AI-generated abstracts and relevance scores

• User feedback integration for continuous improvement

KPIs and Expected Impact

• ⏳ Up to 75% reduction in time spent on weekly legal research

• 📚 ~60% faster access to relevant regulations

• 📈 ~45% increase in coverage and classification of regulatory topics

• 🔁 Continuous feedback loop for ongoing model fine-tuning

3. Customer FAQ Chatbot (POC)

Extended Description

An intelligent customer-facing chatbot designed to enhance FAQ navigation on the e-commerce platform by providing real-time answers based on historical tickets and user behavior. The system identifies gaps in the existing knowledge base, recommends potential FAQ expansions (TO-BE content), and delivers contextual responses using natural language generation. The primary goal is to reduce customer service volume while improving satisfaction through self-service efficiency.

Key Components

• Ticket analysis engine to extract recurring questions

• FAQ expansion and enrichment logic

• Natural language answer generation

• Frontend chatbot interface

KPIs and Expected Impact

• 📉 40–50% reduction in level-1 customer service tickets

• 🔄 Monthly enrichment cycles for FAQ coverage

• 📊 ~30% increase in customer self-service completion rate

• ⭐ Net Promoter Score improvement by 10–15 points

4. Internal Knowledge Base Chatbot (POC)

Extended Description

A conversational assistant aimed at supporting onboarding and training for new hires across operational teams. The chatbot enables fast, targeted access to manuals, procedures, and internal knowledge assets through natural language queries. Designed especially for large-scale onboarding cycles, the solution reduces reliance on tutors and enhances the training process with accessible, always-available guidance.

Key Components

• Semantic search engine over internal manuals and procedures

• User-friendly frontend interface

• Indexed datasets from internal documentation systems

KPIs and Expected Impact

• 👥 50–60% fewer questions routed to human tutors

• ⌛️ 30% reduction in time to onboarding completion

• 📖 70% increase in usage of internal knowledge assets

• 🧩 Scalable to additional business units beyond initial onboarding scope

5. Ticket Categorization and Knowledge Base Foundation (Planned)

Extended Description

This initiative aims to standardize and unify the classification of customer service tickets across channels such as Salesforce, Jira, and email. By clustering recurring issues and aligning metadata, the project will create a foundation for automation and advanced analytics. Incoming email data is treated as a “golden dataset,” driving the creation of a shared knowledge base for first-level automation and future chatbot integration.

Key Components

• Cross-platform classification and clustering model

• Unified ticket ontology aligned across departments

• Golden dataset extraction from historical email flows

KPIs and Expected Impact

• 🧱 90%+ classification coverage of top ticket categories

• 📌 85%+ initial classification accuracy

• ⚙️ Enablement of automation on ~35% of recurring ticket flows

• 🧾 70% improvement in visibility of volume, trends, and urgency across issues

6. Service Desk Assistant with Jira Integration (POC)

Extended Description

A modular set of AI assistants embedded within the service desk environment to improve operational efficiency. These include: a support chatbot for internal users, a helpdesk assistant for frontline service agents, and a transactional agent capable of executing actions on Jira tickets (e.g. assignments, closures). Initial testing focuses on integration with both Jira Cloud and Data Center environments to ensure reliable data ingestion and real-time action capability.

Key Components

• Custom Jira connector for cloud and on-prem environments

• Knowledge base integration and retrieval system

• Semi-automated action engine (e.g. escalate, close, classify)

KPIs and Expected Impact

• 📈 45–60% automation of first-line responses

• 📊 Real-time detection of trending issues or anomalies

• ⚡️ ~35% reduction in average ticket resolution time

• 🎯 Enhanced efficiency and training for junior service agents (up to +40%)

Retail and Consumer Goods / AI for Internal Optimization and Consumer Engagement

Major Retailer (2025) – AI Categorization for a Retailer’s Legal Office

Automatically retrieves new laws and regulations (e.g., Eurolex, Official Gazette) and categorizes them based on relevance to the retail industry, significantly reducing manual analysis efforts.

• Technologies: Vertex AI, Gemini

• Results: 60% reduction in legal analysis time.

Major Retailer (2024) – Demand Forecasting for Retailer Promotional Campaigns

Predicts optimal stock levels during marketing campaigns to reduce waste and maximize sales, improving supply chain efficiency.

• Technologies: Google BigQuery ML (BQML), Vertex AI

• Results: Increased stock forecasting accuracy by 20%, reduced overstocking costs.

Major Retailer (2024) – Generative AI for Internal Workflows and Consumer Engagement

Launched five AI streams on Google Cloud (Gemini) to enhance internal operations, including legal team workflows and consumer engagement. Generative AI was also utilized to assist in inventory management and customer service improvements.

• Technologies: Google Cloud (Gemini)

Luxury and Fashion / AI for Global Campaigns and Product Development

International Luxury Company (2024) – Generative AI & Translations for Global Marketing Campaigns

Utilized Google Cloud’s Vertex AI for generative AI models to support global marketing efforts. This included AI-powered translations, enabling seamless multilingual campaigns across different regions.

• Technologies: Vertex AI, Generative AI

Luxury Jewelry Brand (2024) – AI for Jewelry Design and Cataloging

Leveraged generative AI to assist in modifying existing jewelry designs and generating new sketches. Additionally, AI was used for cataloging 20 years’ worth of design sketches, improving organization and accessibility.

• Technologies: Vertex AI, AutoML

Manufacturing and Industrial Applications / AI for Quality Control and Sales Optimization

Industrial Manufacturer (2024) – AI for Commercial Offer Ranking and Generation

Implemented data science and generative AI models to rank and generate commercial offers, streamlining the company’s sales and offer management processes.

• Technologies: Generative AI, Data Science Models

Industrial Manufacturer (2024) – Visual Inspection on an Assembly Line

Developed an AI-based visual inspection system to identify faulty cable connections or missing components at the end of the production line, improving product quality and reducing defects.

• Technologies: YOLO, Vertex AI

• Results: 30% reduction in assembly errors, improved final product quality.

Healthcare and Sports Analytics / AI for Risk Mitigation and Compliance

Healthcare Tech Company (2024) – AI for Spillage Detection in Lab Samples

• Developed a system using Google Cloud’s Vertex AI to detect spillage in test tube samples within video footage, improving safety and operational efficiency in laboratory environments.

• Technologies: Vertex AI, Cloud Vision, Object detection, Cloud Run, Streamlit, Python

Sports Club (2024) – Injury Prevention App for Professional Football Players

Designed a predictive model using Google Cloud and Vertex AI to forecast injury risks for athletes (soccer player) based on biomedical data. This model enhanced the club’s ability to monitor player health and prevent injuries before they occurred.

• Technologies: Vertex AI, AutoML, Looker (external team)

• Results: 25% reduction in preventable injuries.

Sports Event Organizer (2024) – AI for Brand Recognition in Sports Events

Developed an AI model to categorize and recognize running shoe brands worn by athletes during sports events, allowing better tracking of brand exposure for sponsorship analytics.

Technologies: YOLO, Gemini

Results: Automated identification with 95% accuracy, increased value for sponsors.

Media and Entertainment / AI for Talent Matching and Content Moderation

Media Platform (2024) – AI for Talent Matching and Content Moderation

Built a platform that matches clients with actors and entertainment professionals through an online showcase. Additionally, implemented machine learning models to enhance content moderation, ensuring a safer and more efficient user experience.

Technologies: Machine Learning, AI Content Moderation

Customer Support and Travel / AI for Chatbot Automation and Virtual Assistance

Various Clients (2024) – AI-powered Chatbots and Advanced Search Solutions

Developed AI-powered chatbots and search engines for various applications, including:

• Customer Support: Intelligent chatbots for 24/7 automated assistance.

• New Employee Support: AI-driven FAQ management and onboarding.

• Travel Planning: Advanced virtual assistants for personalized travel experiences.

Technologies: Vertex AI, Conversational Agents

Results: 40% reduction in response times, increased user satisfaction.

Pre 2024 Projects

Falck Renewables (Pre-2024) – Predictive Maintenance for Wind Turbines

Developed a machine learning system on AWS to predict maintenance needs for a wind farm consisting of 50 wind turbines. The project improved operational efficiency and reduced downtime by predicting potential failures before they occurred.

Technologies: AWS, Predictive Maintenance AI

Automotive Company (Pre-2024) – ML Engineering for Data Science Team

Designed machine learning pipelines to support a team of 30 data scientists, improving their ability to deploy models and manage data workflows efficiently.

Technologies: Machine Learning, Data Pipelines

Postal Services Company (Pre-2024) – Hierarchical Cash Flow Forecasting

Developed a predictive model for hierarchical cash flow forecasting, improving financial planning and resource allocation processes.

Technologies: Predictive Analytics, Data Science Models