A handful of projects
I'm proud of.
Across enterprise software, autonomous agents, and open-source tools. Some shipped to thousands of users, some are still rough drafts.
Investi
An autonomous trading desk built from three agents that disagree productively. Researcher, Analyst, Trader—each with veto power.
OpenModel
100,000 simulated consumers, run before a single dollar of marketing spend. A market simulation framework for testing policy, pricing, and messaging.
Motorola · Edge AI Voice Tagging
Shipped voice tagging on body-worn cameras to 20+ enterprise police departments. 75% less data entry, even when the radio drops.
Explicit Agent
Every tool call. Every state change. Visible. A minimal agent framework that refuses to hide what it is doing.
Outcomes, not output.
- $50M ARR retained at Motorola through a 100+ customer migration Enterprise · 90% retention
- 75% Reduction in officer data-entry time with Edge AI voice tagging Voice UX · Offline-first
- 100K Synthetic consumers simulated before a marketing dollar moved OpenModel · LLM personas
- 95% Accuracy on a human–robot collaboration task with multi-agent RL Research · MAGICS Lab
I'm Gabri—an Italian engineer who took the long way around: Groningen, Delft, Boston, San Francisco, now mostly between SF and Amsterdam.
I spend my days thinking about how to ship AI products that hold up under real-world use—and my evenings building small ones to find out what that actually takes.
A few things I keep coming back to when building with AI.
- 01
Ship.
Frontier models are easy to demo and hard to ship. I try to spend most of my time in the boring middle—evals, edge cases, latency, fallbacks—because that's usually where AI products earn (or lose) their keep.
- 02
Trust.
Trust feels less like a feature and more like a side effect of small, honest decisions. I try to design for the moment someone asks "why did it do that?"—and make sure there's a real answer nearby.
- 03
Transparency.
Black-box agents tend to be a short-term win. Where I can, I lean toward systems that expose their reasoning, tools, and state—so teams can debug them, users can correct them, and I can sleep at night.
From research labs to product orgs.
- 2024 — 2026 San Francisco
Product Manager · Body-Worn Cameras
Motorola Solutions
- Led 20+ enterprise deals ($1M–$20M each) across local and federal agencies, owning workflow, customization, and security requirements.
- Shipped Edge AI voice tagging—cut officer data entry time by 75% (20s → 5s) with offline-first VUI.
- Retained $50M ARR through a 100+ customer End-of-Life migration with 90% retention.
- Integrated C2PA standards to harden the product against deepfake manipulation.
- 2023 Santa Clara
Product Manager Intern
Sensitel
- Validated $5M B2B asset-tracking opportunity in agriculture from 20+ supply-chain interviews.
- Designed UX mockups translating raw research into a shippable MVP.
- 2021 — 2023 Boston
AI Research Engineer
MAGICS Lab · Northeastern
- Built an open-source framework for multi-robot cooperative simulation.
- 95% accuracy on a human–robot collaboration task using multi-agent RL.
- MSc Engineering Management Northeastern University Summa Cum Laude · 4.0
- MSc Mechanical Engineering TU Delft Cum Laude
- BSc Industrial Engineering University of Groningen Cum Laude
Get in touch
I'm always up for a good
conversation about applied AI.
Working on something interesting in AI products? Whether it's a question, a collaboration, or just a chat—I'd love to hear about it.