RESUME

ZÜRICH, SWITZERLAND SWISS_C_PERMIT 9+_YEARS SHIPS_TO_PROD

Agents in
production,production,
not decks.

AI TECH LEAD · AGENTIC SYSTEMS & APPLIED AI

9+ years across applied AI, data science, and consulting. I lead agentic AI system design in the AI Center of Excellence of a leading luxury retailer, shipping production AI end-to-end on GCP and owning the company's agentic framework from architecture through dev/prod rollout. Quantitative foundation (MSc Quantitative Finance, MPhil Statistics), plus a parallel track building revenue-generating AI for SMEs as an independent consultant.

SEE_THE_PROOF BOOK_A_CALL
MATTEO_BASSIZÜRICH, SWITZERLAND
€1M/YR SAVED · TRANSLATION AGENT, LIVE SINCE 2024 €1M+ INCREMENTAL SALES · CDP, YEAR-LONG A/B TEST 30K+ IMAGES · 3 MODEL GENERATIONS 8 → 30 ENGINEERS ON AI-ASSISTED ENGINEERING 30 STORES → GLOBAL ROLLOUT · IMAGE RECOGNITION €50K/MO CLIENT PLATFORM · BUILT END-TO-END
01 · SELECTED_WORK

Selected systems and results.

Employer and client names are withheld. These metrics cover deployed systems and current rollouts for a leading luxury retailer, plus independent work for Italian SMEs.

EMPLOYER WITHHELD
AGENTIC_FRAMEWORK · GCP · MCP · GOOGLE_ADK

The Agentic Framework

DEV → PRODGOVERNED PROMOTION PIPELINE, MY DESIGN

The company's agentic framework. Agents compile recurring tasks into deterministic automations on internal MCP servers, growing a reusable automation library. The stack runs hybrid, with citizen-agent tooling for business users alongside custom MCP servers and Google ADK on GCP for the pro-code track, all governed by a dev → staging → prod approval pipeline.

2023 → PRESENTARCHITECTURE → ROLLOUTCOMPANY-WIDE

┌──────────────┐   ┌───────────────────┐
│ CITIZEN      │   │ PRO-CODE TRACK    │
│ AGENT TOOLS  │   │ Google ADK · GCP  │
└──────┬───────┘   └────────┬──────────┘
       │                    │
       ▼                    ▼
  ┌─────────────────────────────┐
  │   INTERNAL MCP SERVERS      │
  │   reusable automation lib   │
  └──────────────┬──────────────┘
                 ▼
   dev ──▶ staging ──▶ prod
   (approval pipeline, my design)
EMPLOYER WITHHELD
TRANSLATION_AGENT

Translation Agent

~€1M/YRSAVED · IN PRODUCTION SINCE 2024

A translation agent for a leading luxury retailer, in production since 2024 and saving about €1M per year.

2024 → PRESENTGCP

EMPLOYER WITHHELD
CUSTOMER_DATA_PLATFORM

Customer Data Platform

€1M+INCREMENTAL SALES · YEAR-LONG A/B TEST

Led the CDP on GCP, with company-wide segmentation deployed internationally, including on-site work in China for data residency. A full year of A/B testing validated €1M+ in incremental sales.

2020 → 2022GCPA/B VALIDATED

IN ROLLOUT
IMAGE_RECOGNITION

Store-Window Vision

30 → ALLSTORES · INITIAL ROLLOUT TO FULL GLOBAL NETWORK

Image-recognition tool for store windows, in rollout across an initial 30 stores and scheduled to extend to the full global store network.

IN ROLLOUTCOMPUTER VISION

3 GENERATIONS
IMAGE_GENERATION

Image-Gen Platform

30K+IMAGES OVER 3 YEARS · 40 USERS SCALING TO 80

The platform spans three model generations: LoRA fine-tunes of SD 1.5, a fine-tuned SDXL kept in production by user demand, and Gemini image models now in rollout.

30K+ IMAGES · 3 YEARSLoRA · SDXL · GEMINI

CLIENT-HOSTED
INDEPENDENT_CONSULTING

Multi-Tenant Platform

€50K/MOREVENUE · DESIGNED & BUILT END-TO-END

For an Italian SME, I designed and built a multi-tenant online platform end-to-end. It is fully client-hosted and generates €50K/month in revenue. AI solutions for a second SME generate €8K+/month for that client.

2019 → PRESENTEND-TO-END

V2 LIVE
AI_ADOPTION · LEAD_GEN

Engineering, Amplified

8 → 30ENGINEERS ON AI-ASSISTED ENGINEERING

Rolled out AI-assisted engineering (Claude Code + best-practice standards) across the company. Also shipped a B2B lead-generation tool, live for a year and now in its second major version.

COMPANY-WIDECLAUDE CODE

02 · EXPERIENCE

Nine years, one direction.

Consulting → analytics → data science → agentic AI leadership. Employer and client names are withheld; roles and approved results are shown instead.

2023 → NOW

AI Tech Lead · Agentic Systems & Applied AI

A LEADING LUXURY RETAILER · ZÜRICH, CH

  • Rolled out the internal AI platform; lead the technical rollout of the agentic framework across a hybrid stack (citizen-agent tooling for business users, plus custom MCP servers and Google ADK on GCP for the pro-code track), owning coded agents end-to-end across architecture, development, deployment, and user management.
  • Translation agent (~€1M/yr saved), image recognition for 30+ stores, image-gen tooling (30K+ images, 3 model generations), B2B lead-gen tool v2.
  • Drove AI-assisted engineering adoption company-wide: 8 → 30 engineers.
  • Company-wide AI advisor; quality-gates external consultant deliverables.
2020 → 2022

Data Scientist / Product Lead

SAME RETAILER · ZÜRICH, CH

  • Led the Customer Data Platform on GCP: €1M+ incremental sales, validated by a year-long A/B test.
  • Built the company-wide customer segmentation model, deployed internationally, including on-site work in China to satisfy data-residency constraints.
  • Designed the customer engagement score as my first data-science project; it remains a key CRM metric five years on. Also owned KPI frameworks, demand forecasting, and production pipelines.
2018 → 2020

Data Analyst

A GLOBAL TRAVEL-TECH GROUP · ZÜRICH, CH

  • Predictive analytics for hotel demand forecasting and dynamic pricing.
  • Interactive Tableau and Power BI dashboards; Snowflake table design during the integration of three merging companies.
2016 → 2018

Tech & Strategy Analyst

A STRATEGY CONSULTING FIRM · MILAN, IT

  • Advised financial-services clients on data strategy and AI-driven automation opportunities.
03 · TECHNICAL_STACK

Tools of the trade.

Core tools across agentic AI, GenAI engineering, cloud, data, and programming.

AGENTIC_AI

  • Google ADK
  • MCP server design & prod standards
  • Multi-agent orchestration
  • Agent evaluation
  • Dev/prod promotion patterns

GENAI_ENGINEERING

  • LoRA fine-tuning (diffusion)
  • SFT + DPO (LLMs)
  • RAG
  • Evals
  • Prompt / context engineering
  • Claude Code
  • Vertex AI
  • Local inference

CLOUD_&_DATA

  • GCP (Vertex AI, Firestore)
  • Data pipelines
  • Power BI / DAX
  • Tableau

PROGRAMMING

  • Python
  • SQL
  • JavaScript / TypeScript
  • React · Next.js
  • Reviewing & hardening AI-assisted codebases

EDU: MPhil Statistics · MSc Quantitative Finance (110/110) · BSc Economics LANG: IT native · EN fluent · DE B1

04 · OFF_THE_CLOCK

Personal projects.

I fine-tune open models on cloud H100s and run quantized models on a self-built eGPU setup. The baby monitor keeps all processing local.

V.01

MoE Fine-Tuning Lab

Two fine-tunes of open-weight MoE models (Gemma family): curated-dataset SFT, then SFT + DPO with agent-generated preference data, trained on cloud H100s.

V.02

Local Inference Rig

Self-built eGPU setup over OCuLink serving quantized open-weight models. I use it to test quantization trade-offs and MoE architectures.

V.03

Privacy-First Baby Monitor

A self-hosted baby monitor with all processing kept local.

05 · CONTACT

Start something.

Book a time on Calendly or connect on LinkedIn. The CV is available from the header and below.

BOOK_A_CALLCalendly · choose a time LINKEDIN/in/matteo-bassi GITHUB/MBassi91 RESUMEcv.pdf · download

// PROPOSAL_DEMO · No message will be sent. Book a call to get in touch.