Nicola Valenti

Product Manager · Co-founder

Two years building AI products from first principles. Now taking them to enterprise pilots.

Co-founder and Head of Product at Nordic Loop. I got into product by starting a company: customer discovery, AI pipelines, data modeling, and enterprise pilots. I grew up in Milan doing photography, which probably explains why I care as much about how something looks as whether it works. Relocating to Stockholm in September 2026 and looking for a PM role.

Selected Projects

Strategy Discovery B2B

01 Customer Discovery

Marketplace Pivot

Led 135 customer conversations that surfaced the real barrier: enterprises couldn't see their own surplus before trying to share it. Repositioned from external marketplace to internal inventory system, cutting the deal cycle from 4–5 months to 2.

Read Case Study ↗

AI MVP Cross-Functional

02 Cross-Functional

AI × CO₂ MVP

Four modules — AI classification, CO₂ estimation, UI, backend — advancing in parallel with no shared integration contract. Defined the minimal workflow, froze the seam at {category, quantity, unit}, and delivered a working internal demo in 2 weeks.

Read Case Study ↗

AI Architecture PRD CSRD

03 AI Architecture

LLM Pipeline

YOLO's classification ceiling was becoming a compliance risk for Digital Product Passports. Scoped a two-model architecture (YOLO + Gemini Flash) that added subcategory and composition detection, cutting manual tagging time from 2 minutes to 30 seconds.

Read Case Study ↗

Systems Thinking Data Model 0→1

04 Systems Thinking

Material → Asset

The material-based taxonomy was misaligned with how the AI classifies, how CO₂ is calculated, and how construction workers think. Redesigned the data foundation before any pilot launched — zero post-launch migrations, CO₂ estimates reliable from day one.

Read Case Study ↗

How I Work

01
Makes decisions in the room. Documents the reasoning after.
02
Finds the line between a product decision and a technical one.
03
Can read a schema. Can't write a migration.
04
Tests for the worst user, not the average one.