Founder of CosentriQ | Building human-centered AI evaluation infrastructure

I build systems that help AI teams understand whether their products will actually work for people in the real world.

My work focuses on the gap between technical model performance and human outcomes: how people understand, trust, adopt, and act on AI-generated outputs.

Because an AI system can produce a technically correct response and still create confusion, inappropriate trust, over-reliance, failed adoption, or poor real-world decisions.

I’m building the evaluation infrastructure to surface those risks before they scale.

CosentriQ combines agentic simulation with structured human validation to evaluate how people understand, trust, adopt, and safely act on AI outputs—surfacing trust, adoption, safety, and decision risks that technical evaluations miss.

The platform helps AI teams evaluate the human side of model behavior across different users, scenarios, and real-world contexts from pilot to production.

CosentriQ turns those signals into structured evaluation sprints that help teams refine model behavior before launching, expanding, or scaling their AI products.

DollarFifteen is the paid contributor network powering CosentriQ’s human validation layer.

Contributors evaluate AI outputs, interactions, and real-world scenarios through structured, mobile-first tasks. Their judgment helps surface behavioral patterns, interpretation gaps, and real-world risks that technical evaluations and agentic simulations may not fully capture.

D15 is designed to make human evaluation more representative, economically participatory, and operationally useful for AI development.

I am developing an evaluation approach that compares predicted human behavior from agentic simulations with observed human responses.

The gap between those signals helps teams identify where simulated expectations diverge from real-world interpretation, trust, behavior, and decision-making.

This is the foundation of CosentriQ’s Signal Gap intelligence.

Understand the users, decisions, incentives, constraints, and failure conditions before building the interface.

Surface structural, behavioral, safety, and human risks before the market discovers them.

An AI output must not only be accurate. People must also be able to understand it, trust it appropriately, and act on it safely.

Agentic simulations can predict how different users may respond. Human validation reveals what actually happens.

Both signals matter. The gap between them is valuable intelligence.

Every evaluation should strengthen the system, improve future decisions, and deepen understanding over time.

• Second-time founder
• Founder and CEO of CosentriQ
• Previously built DivySci into a multimillion-dollar company serving more than 10,000 users
• M.S. in Information and Knowledge Strategy, Columbia University
• B.S. in Computer Science, Pace University
• Two-time NSF-funded founder
• Previously supported by Google for Startups, AWS, Camelback Ventures, Black Ambition, and the Roddenberry Foundation

AI and Evaluation

LLMs · RAG · Agentic Systems · Human Evaluation · AI Safety · Model Behavior · Multi-Agent Workflows

Backend

Python · Node.js · PostgreSQL · Supabase

Frontend

React · TypeScript · Next.js

Product and Systems

AI Evaluation Infrastructure · Human Signal Systems · AI Product Strategy · Decision Systems · Product Architecture

Human-Centered AI Evaluation · Human × Agentic Evaluation · AI Safety · Trust and Decision Risk · Human Signal Infrastructure · Responsible Model Behavior

The sandbox works. The real world is messy.

I build the layer that helps AI teams understand the difference.

Website: www.arianaabramson.com

LinkedIn: linkedin.com/in/arianaabramson

Email: ariana.abramson@gmail.com