About
About
The Short Version
I’m Lucas Cazelli, CPO & Co-founder at North AI, where we’re building synthetic audiences for video analytics using neuroscience-inspired AI.
Previously, I optimized physical structures as a civil engineer. Now I optimize attention and decision-making in digital products.
Based in London. Building in public. Writing about the intersection of neuroscience, product management, and statistical thinking.
The Longer Version
From Structures to Synapses
My journey started in civil engineering, where I spent years applying finite element analysis to optimize load distribution in physical structures. The transition to product might seem unusual, but the core challenge remains the same: understanding how complex systems handle stress and distribute resources.
In engineering, it was about concrete and steel. In product, it’s about cognitive load and attention.
What I’m Building Now
At North AI, we’re creating what doesn’t exist yet: synthetic audiences that can predict how real humans will respond to video content. We combine:
- Eye-tracking and gaze synchronization patterns
- EEG-proxy metrics for cognitive engagement
- Multi-agent AI systems for behavioral simulation
- Statistical clustering for psychographic segmentation
Think of it as a “virtual focus group” that never gets tired, never lies, and can represent millions of viewers simultaneously.
The Builder Mindset
I’ve always been a maker. Whether it’s:
- Software: From accounting APIs processing 50M daily transactions to real-time video analytics
- Hardware: Arduino-powered kinetic sculptures and 3D-printed mechanisms
- Products: Launching everything from blockchain film financing to B2C streaming platforms
The tools change. The drive to build doesn’t.
My Approach to Product
Discovery ≠ Delivery. I run dual-track product development with clear gates between exploration and execution.
I believe in:
- Outcome-driven roadmaps over feature factories
- Statistical rigor over gut feelings (though System 1 thinking has its place)
- Rapid experimentation with <30-second calibration flows, not 2-minute ones
- Clear communication - if it can’t fit in a one-pager, it’s not ready
Technical Foundation
- Engineering: BSc Civil Engineering, structural optimization
- Data: Advanced statistics, Six Sigma, experiment design
- Product: 5+ years leading products from zero to exit
- Code: Still ship production code (Python, JS, SQL)
- AI/ML: RAG architectures, multi-modal pipelines, synthetic data generation
What I’m Learning
Currently diving deep into:
- Neuroscience applications in UX (beyond “dopamine hits”)
- Bayesian approaches to audience prediction
- Generative agents for behavioral simulation
- The cognitive science behind creative decision-making
Outside Work
- Fitness: Muay Thai, BJJ, and arguing with my pull-up bar
- Making: Building unnecessary robots that my wife tolerates
- Gaming: Optimizing Baldur’s Gate 3 builds with the same rigor I apply to product metrics
- Languages: Fluent in English and Portuguese, functional in Italian and Spanish, learning German one phrase at a time
Let’s Connect
I write about product, neuroscience, and decision-making here and across:
- Professional: LinkedIn | GitHub
- Thoughts: X/Twitter | Bluesky
- Writing: Medium | Blog
- Email: lucascazelli@gmail.com
What I’m Available For
- Speaking: Neuroscience of attention, synthetic audiences, product analytics
- Advisory: Early-stage deep tech, video analytics, B2C products
- Collaboration: Research projects combining AI and cognitive science
- Coffee: Always, if you’re in London
Currently: Building the future of video testing at North AI
Previously: Made buildings stand up, then made products take off
Always: Asking “what does the data actually say?”