Neuroscience → AI systems
I build AI systems that think like teams, act like operators.
I study how people make decisions at the neural level — reward, motivation, cognitive bias, risk tolerance — then build systems that account for those realities. Production-deployed, not notebooks.
Featured work
Live systems with real stakes — each one its own space. Scroll through; every frame links to source.
NovaCRM
Turns inbox chaos into a sales pipeline
A live AI-native CRM where six agents read scattered email and Slack and turn them into structured, prioritized deals, with a human in the loop for the calls that matter.
TRIBE v2
Predicts how a video lands in the brain
It scores short-form video by predicted brain response, modeled from the video alone with no scanner, and commits its success criteria to git before the results, so the claim can be proven wrong.
Sector Flow Analyzer
Catches sector rotation before it hits price
It detects institutional money rotating between market sectors in the covariance structure before the move shows up in price. Decision-support only.
Spots regime shifts early, and it's been tested honestly.
p = 3.26e-20
the signal is real, not random noise
13 yr
validated on unseen, out-of-sample data
60.7%
calls right — vs a 55% coin-flip
alfred-v2
My always-on second brain
A self-hosted memory system that captures everything I read, answers in under a second, and repairs itself before it ever pages me.
Capture → recall in under a second → self-repair
Obsidian vaults
everything I read · 6 domains
pgvector graph
30k+ records, embedded
MCP query bridge
answers in under a second
self-healing watchdog
detect → heal → page (last resort)
Sentiment models, automation infra, and the experiments that don’t make the front page. It’s all on GitHub.
All repositories ↗About
I build agentic AI systems: software where autonomous agents read through messy human information, reason about it, and carry the work all the way to action. I come at it from an unusual angle. I study how people actually decide (reward, motivation, bias, at the neural level) and design systems with those realities built in. Builder first; the science is the lens I build through.
My main system, NovaCRM, is a live AI-native CRM where six specialized agents turn scattered email and Slack into a structured pipeline. Each agent has one bounded, verifiable job, and a person stays in the loop for the moments that matter. It grew out of Executive Mind Matrix, where three agents with competing cognitive biases argue a decision before routing it. Around them I’ve built a knowledge system that self-heals before it pages me, a model that scores video by predicted brain response (success criteria committed to git before the results), and a behavioral-finance tool that catches institutional rotation in covariance before it shows up in price.
The thread through all of it: the bottleneck is rarely the model. It’s the system around it, and the people it’s for. I care about the gap between good thinking and executed action, because I’ve watched capable people drown in operational overhead while their best ideas never ship. So I’m putting real agentic systems into production and learning in the open, headed toward an early-stage team where building the system and understanding the people it serves are the same job.
- Domain
Agentic AI systems, grounded in the neuroscience of decision-making
- Builds
NovaCRM · alfred-v2 · TRIBE v2 · Sector Flow · Executive Mind Matrix
- Studies
Psychology & Entrepreneurship · cognitive science w/ a computational-neuroscience grounding
- Direction
AI product / engineering at an early-stage team
Work with me
I help teams put agentic AI into production — the system around the model, built to ship and to hold up once real people depend on it.
Most AI projects stall in the gap between a promising demo and something an operator can trust on a Monday morning. That gap — ingestion, grounding, verification, a human in the loop for the calls that matter — is the work I do. I've built it for my own systems; I build it with teams who need it for theirs.
- Agentic systems
Multi-agent pipelines where each agent has one bounded, verifiable job — not a single prompt hoping for the best.
- Decision intelligence
Turning messy human information — email, docs, signals — into a prioritized, explainable next action.
- Production hardening
Grounding, evals, and human-in-the-loop layers so a system fails loud, self-heals, and earns trust.
- Founders and teams with an AI demo that now has to become something operators rely on.
- Operators drowning in manual triage that agents could be doing — correctly and verifiably.
- Anyone who needs the system around the model built right, not just the model called.
Have a system that needs to make it to production?
Tell me the problem in a paragraph. If it's a fit, we'll find a time to talk.