Pareto.AI

We've been alerted to scammers posing as Pareto.AI team members using fraudulent email addresses to offer fake work opportunities and steal sensitive information. Please note: Official communications from Pareto.AI will *only* come from an @pareto.ai email address. If you receive a message from a similar but incorrect domain, it is a scam. If in doubt, please feel free to contact us through our official channels. We are actively working to report and shut down these fraudulent accounts, so if you’ve been targeted, we'd appreciate it if you reported it to us.

Last week at Pareto.AI, we hosted an RL Researcher Roundtable and dove into the practical challenges of training RL agents in production, with invaluable insights from our panelists - Danielle Perszyk (Cognitive Scientist @ Amazon AGI), Alex Gajewski (founded SF Compute), and Kyle Vedder (Researcher @ Dyna Robotics). Researchers from Physical Intelligence, x.AI, Google DeepMind, Meta SuperIntelligence Labs, UC Berkeley, and more joined the discussion. Thanks Sylvia Tong and Cai Chen for collaborating on this. At Pareto.AI, we’re building a tight, high-signal community of AI researchers and domain experts in fields ranging from molecular biology to law - united in advancing the frontiers of AGI. If you’d like to join us at future sessions, please keep an eye for future events on our  Luma calendar here: https://luma.com/pareto.ai :))

A week in the Philippines with our Pareto community - a powerful reminder that the future of work is borderless, human-first, and deeply connected. What happens when brilliant minds from Chile, India, Kenya, South Africa, the USA, the Philippines, Spain, and Colombia come together? You get eight countries, four continents, and one unstoppable force of energy, curiosity, and heart. Our offsite wasn’t just a gathering; it was a living example of our core belief: that work should amplify our humanity, not replace it. From strategy huddles to karaoke nights, every moment reflected the Pareto DNA - connection, co-creation, and pushing for excellence together. As we grow and shape what’s next, these shared experiences remind us why we’re here: to build a world where experts, operators, and dreamers can do their best work - no matter where they’re from. Here’s a glimpse into the spirit that powers us forward. #paretoai #futureofwork #humaneintelligence #aiforhumans

**WARNING ABOUT SCAM ATTEMPTS: Any email or message from a Pareto.AI employee would come from an official *****@pareto.ai email address. If you receive an email from someone pretending to be me or a Pareto.AI employee please double check that the message comes from the correct pareto.ai suffix** Dear friends and others in my network: It enrages me that I have to make this notice, but there are scammers out there pretending to be me or another member of Pareto.AI. While we are actively recruiting for all kinds of project work opportunities all across the globe, there are people out there fraudulently attempting to scam you. Please double check the email address of any messages being sent to you that it comes from an official pareto.ai email.

We’re thrilled to welcome Ray Phelps as our new Data Delivery Lead at Pareto! Ray brings a wealth of experience from their time as a Data Scientist at the NYC Campaign Finance Board, where they analyzed campaign finance data to help reduce the influence of money in local elections. Their expertise in public sector analytics and AI-driven research has already been invaluable to our data operations team. Ray first joined Pareto as an AI trainer, quickly connecting with our mission to shape AI’s societal impact by ensuring human insight guides the development of ethical and aligned AI systems. In their new role, they remain dedicated to advancing this vision, and we're proud to have such a passionate and driven leader on our team. Welcome aboard, Ray! 🚀

A reputed AI lab needed high-stakes evaluation data—precise, instruction-sensitive, and scalable. For months, they relied on Pareto.AI to deliver it. But when budget pressures mounted, they switched to a cheaper vendor—a startup pivoting into data labeling—that promised the same throughput at a lower cost. What they got instead was a disaster. 1. No QA process: Errors went unchecked, corrupting evaluation benchmarks. 2. Speed over quality: Speed is important, but not at the cost of quality! Even though throughput goals were met, labelers failed to follow instructions. 3. No alignment mechanisms: Without structured feedback loops, errors compounded. The results? Months of wasted effort, unreliable model outputs, and research gridlock. When the cost of poor data became unbearable, they came back to Pareto. Our expert-led workflows, multi-layered quality controls, and continuous alignment systems gave them what no cut-rate vendor could: high-quality data that accelerates AI research *reliably.* The best labs in the world don’t gamble with quality. They choose Pareto. Read more in our latest blog below 🚀

For today’s #EmployeeSpotlight, we’re thrilled to welcome Jorge M. Cano as our new Regional Manager for Latin America! 🎉 With a background in industrial engineering and an MBA in international business administration, Jorge brings a wealth of expertise in project management, operations, and AI-driven initiatives. Jorge has led high-impact teams at Amazon, OCG Group, and 3M, driving innovation across machine learning, cybersecurity, and biomedical industries. Outside of work, Jorge is a dedicated dad to a 12-year-old son, balancing his professional duties with the most important job of all—being an awesome parent. Welcome to the team, Jorge! 💜

In AI research, small deviations in ground truth can have outsized consequences. Misinterpretation at the annotation level introduces noise that cascades throughout a model’s lifecycle—compounding errors, distorting evaluations, and ultimately eroding trust in model performance. This is especially dangerous in high-stakes areas like safety and alignment, where edge cases determine success. Over the past year, we have partnered with research teams to analyze and refine their annotation rubrics, systematically identifying edge-case ambiguity and iterating together to make improvements. How can we standardize best practices for evaluation rubrics in a way that scales across models and domains? Can we develop open benchmarks for rubric ambiguity, much like we’ve done for bias or robustness? We dive into the consequences of ambiguity in data annotation rubrics in our latest blog, also explaining how how refining these systems is essential for creating more robust AI models. If you’re thinking about improving your rubric design, we’re happy to collaborate and share what we’ve learned in greater detail!

We’re thrilled to announce that Dr. Mark Whiting has joined Pareto.AI as Chief Technology Officer! Dr. Whiting was previously a Senior Computational Social Scientist at the Computational Social Science Lab at Penn, where he spent his career building systems to study how people behave and coordinate at scale. With a wealth of research, groundbreaking contributions, and a deep commitment to understanding human intelligence, he has emerged as a true leader in this field. At Pareto.AI, we’re building a future where we amplify humane intelligence by enabling AI systems to align more closely with human values. Dr. Whiting’s expertise will strengthen our ability to design smarter, more ethical systems that empower subject-matter experts and shape the next generation of AI. We’re honored to welcome such an accomplished leader to our team and can’t wait to collaborate on the challenges ahead. Humanity to infinity!