TONY MARKSI build AI products that work in the real world.
I'm Tony Marks, an AI founder and product builder. Since 2017, I've designed and deployed conversational, voice, search, and generative AI products for the Olympic Games, professional sports leagues, airlines, retailers, and enterprise teams.
Today, I'm building Hotline Sports and Project Sapiens—and partnering selectively with organizations that need experienced help turning AI ideas into production software.
Clients and Partners
Work
AI products shipped at global scale.
My work spans conversational interfaces, voice AI, intelligent search, data systems, and generative AI. These are selected examples of systems designed for real users, operational constraints, and high-profile environments.
Olympic Games
Official AI experiences across three Olympic Games
Helped design and deploy AI-powered experiences for Rio 2016, PyeongChang 2018, and Tokyo 2020, including work supporting Paralympic and Youth Olympic events.
Professional Sports
Voice AI for the NBA and NFL
Built voice and conversational AI capabilities that helped professional sports organizations make information and digital experiences accessible through natural-language interfaces.
Airlines
Conversational systems for global travel
Designed AI assistants and customer-facing experiences that helped travelers find information and complete common tasks across airline digital channels.
Retail and Live Events
Interactive AI for New Balance
Developed an in-store digital experience for New Balance’s Boston flagship around the Boston Marathon, connecting product discovery, brand storytelling, and customer interaction.
Current Exploration
Pronunciation Intelligence
One problem has stayed with me for years.
Artificial intelligence has become remarkably good at generating language, but it still struggles with one of the most human parts of communication:
People’s names.
While building voice AI for professional sports, I kept running into the same problem.
AI systems frequently mispronounced athlete names—especially names from diverse cultures and languages.
Correcting those pronunciations wasn’t simply a matter of changing one word.
It required verifying how the individual actually pronounced their name, identifying the correct phoneme sequence, and then teaching speech systems how to reproduce it consistently.
That process inspired an ongoing research project:
Building a verified pronunciation database that pairs names with human-verified audio pronunciations and phoneme mappings.
The goal isn’t simply better speech synthesis.
It’s making AI more respectful, more inclusive, and more accurate.
Today I’m continuing to explore pronunciation intelligence across sports, media, voice assistants, and conversational AI.
Research Highlights
- Verified pronunciation database for diverse names
- Human-verified audio pronunciation references
- Phoneme mappings for speech synthesis
- Pronunciation APIs for developers
- Search tools for journalists and broadcasters
- Voice AI research for sports and media
“The smallest details often determine whether AI feels intelligent—or human.”
The original concept grew out of work building conversational AI for professional sports, where accurate pronunciation became essential for athlete introductions, broadcasting, and voice interfaces.
Research Philosophy
Most AI systems guess.
I’m interested in helping them know.
Field Notes
Observations from building at the edge of AI.
Notes on artificial intelligence, product design, and the technologies reshaping how we work and think.
Capabilities
From AI concept to production system.
I work with founders, product teams, and established organizations that need more than a prototype. Engagements can cover strategy, product definition, architecture, implementation, evaluation, and launch.
AI product strategy
Identify the right use case, define the product, select the technology, and establish a practical path to launch.
AI agents and workflows
Design agentic systems that can reason across tools, data, business rules, and human approval steps.
Conversational and voice AI
Build natural-language experiences for customer support, discovery, transactions, internal operations, and connected devices.
Enterprise knowledge and search
Turn fragmented documents, databases, and internal knowledge into useful, governed AI experiences.
Rapid product development
Move from concept to working software using modern models, APIs, evaluation methods, and production-ready architecture.
AI brand visibility
As more people ask AI for recommendations instead of searching the web, companies need to understand how AI systems discover, interpret, and represent their brand.
Read Field Note 002Ventures
What I'm building today.
I continue to build my own products while working selectively with companies on ambitious AI initiatives.
Project Sapiens
Building nowAn applied AI company focused on practical tools, education, and custom systems that help people and organizations work more effectively in an AI-first world.
Project Sapiens is also the operating company through which select consulting and product-development engagements are delivered. (Formerly Sapien Ai)
Hotline Sports
Building nowA modern sports intelligence platform built around one deceptively difficult question: Can I watch this game with the services I already have?
Hotline Sports combines schedules, broadcast availability, location, subscriptions, and team preferences to create a personalized guide to what is on and where to watch.
About
Building production AI since 2017.
I founded Sapien AI in 2017, during the early commercial wave of conversational artificial intelligence. Long before generative AI became a mainstream category, my work focused on a practical challenge: turning emerging AI technology into products that ordinary people could actually use.
Over the following years, I designed and shipped conversational assistants, voice interfaces, intelligent search products, data systems, and custom AI applications across sports, travel, retail, and enterprise software.
That work included AI experiences connected to the Rio 2016, PyeongChang 2018, and Tokyo 2020 Olympic Games; voice AI capabilities for professional sports; customer experiences for major airlines; and an interactive retail application for New Balance.
In 2021, my team participated in OpenAI's developer beta and began exploring early generative models, speech systems, and image-generation technology for enterprise applications.
The Sapien AI domain and brand were acquired in 2025. Today, I am building Hotline Sports and Project Sapiens while working selectively with organizations on AI strategy, product development, and production deployment.
My focus remains the same: identify a meaningful problem, choose the right technology, and build something dependable enough to be used outside a demo.
Contact
Have an AI product worth building?
I work selectively with companies that have a meaningful problem, access to the right data or users, and the ambition to build a real product—not another AI demonstration.
Tell me what you are trying to create, where the project stands today, and what success would look like.
Consulting and product-development engagements are delivered through Project Sapiens, Inc.