Sinan Ozdemir is Head of AI Developer Education at Fireworks AI. He has written about ten books on machine learning and large language models, founded and sold an AI company, and taught in middle schools, universities and Maryland prisons. He was working with language models more than a decade ago, before the transformer existed.
Every one of those jobs has had the same goal: teach people well enough that they stop needing the teacher. Sinan puts it simply: "my job is for you not to need me anymore."
He joined AJ Green on The AI Collective Podcast, and he is exactly the kind of person we built the Hall of Fame for.

Start with the definitions
Sinan trained as a pure mathematician at Johns Hopkins, where he earned his master's and later lectured in data science. His research in algebraic geometry explored elliptic curve encryption for ultra-secure voting machines. The paper ended with an honest verdict: the math worked, but no one could afford to build it, maybe for another 30 years.
Mathematics taught him a habit he never dropped. You define your terms before you argue about anything. Math textbooks make you read the definitions before you reach the first theorem, and Sinan still works that way.
That habit is rare in AI, where words like "agent," "post-training" and "frontier" change meaning every year. It is also why people trust him to explain the field. He does not chase the newest term. He asks what it actually means first.
After Hopkins, Sinan moved back and forth between building and teaching. He founded Kylie.ai, a Y Combinator-backed conversational AI company, which was acquired in 2019. He wrote Principles of Data Science, Quick Start Guide to Large Language Models and Building Agentic AI, taught courses for Pearson and O'Reilly, and consulted for companies like Adobe.
On paper, that looks like four careers. In practice it was one job done four ways. Founding a company meant understanding the technology well enough to ship it. Writing a book meant understanding it well enough to explain it to strangers. Teaching meant making sure the explanation stuck.
Developer education at Fireworks pulls all of it together. Fireworks builds infrastructure for running and customizing open models, and the field moves weekly. Someone has to help developers keep up without drowning, and Sinan had spent his whole career getting ready for that.

Teaching people to make their own maps
Most developer education stops at how to call the API. Sinan's version goes further: giving people a way to keep up on their own.
His method starts with what he calls evergreen pillars, the parts of AI that do not change from release to release. What does a parameter count actually tell you? What does a benchmark measure, and what does it miss? Once people understand those, he helps them map each pillar to what they need. A model that tops a coding leaderboard is irrelevant if you plan to use it for slide decks or marketing copy.
The point is to make people harder to fool, including by the companies building the models. As Sinan puts it, "Don't let the organization making the model tell you what you care about."
On his second screen he keeps a chart of every model on Fireworks, plotted by size against intelligence. The ones that interest him are small models that score unusually high. Intelligence per parameter is the question he keeps coming back to, because it points toward a future where anyone can run a capable model on their own laptop.

Past the next coding model
Sinan wants the industry to widen its idea of progress. Coding models are valuable and well funded, but new frontiers are thin: there are still no serious foundation models or evaluations for work like marketing or sales. When Adobe asked him which model was best for marketing, he had no answer, because no one had defined what good marketing output looks like.
He pushes on what counts as valuable at all. OpenAI's charter defines AGI around economically valuable work, and Sinan asks who gets to decide what that means. An AI that helps someone finally learn a video game they always wanted to play will never show up in GDP. It would still make their life better.
"I have to sit here and teach like a teacher might."
Writing for a field that keeps moving
Sinan's newest book, written with Fireworks, came out the day before he joined the podcast. Writing it meant committing definitions to paper in a field that refuses to hold still, knowing someone will point back at the page in eight years and say it aged badly.
His answer is to define things at the level that lasts. Ask him what an agent is and he skips "an LLM with tools." To him it is a model that can reach beyond its own parameters. Tools are one way to do that today, and the definition still holds if tools change.
He applies the same thinking to people. The industry talks constantly about recursive self-improvement for AI. Sinan's question is what happened to recursive self-improvement for humans. That, he points out, is what education has always been.
North star: his students
Sinan does not measure his career in books. The books, videos and courses are the byproduct. What matters is the people they reach: former students who are now VC partners, math teachers and AI teachers, passing the work on without him.
"My students are the only people that matter to me," he says.
That belief sits at the center of The AI Collective too. Nobody keeps up with AI alone, and the goal is a community where people learn to think for themselves about what this technology means for their lives.
A Seat in the Hall
He took a mathematician's love of definitions into a field that keeps rewriting them. He built and sold a company, wrote the books a generation of practitioners learned from, and now teaches developers to judge models for themselves. He counts his success by how little his students need him.
Sinan Ozdemir has earned his place in our Hall of Fame.
Find Sinan at sinanozdemir.ai, on LinkedIn, and as @prof_oz on X.
The AI Collective is built by volunteers across 180+ chapters in 40 countries.
Thank you to the thousands of volunteers around the world who make this work possible. We truly could not do this without you.
🧑💻 About the Editorial Team

About Josh Evans
Josh is a Managing Editor at The AI Collective Newsletter and leads content for The Byte. Outside of AIC, Josh works in Content Protection at Spotify.
About Our Hall of Fame Inductee

About Sinan Ozdemir
Sinan Ozdemir is Head of AI Developer Education at Fireworks AI and the author of roughly ten books on machine learning and large language models, including Principles of Data Science and Quick Start Guide to Large Language Models. A former Johns Hopkins lecturer with a master's in pure mathematics, he founded Kylie.ai, a Y Combinator-backed conversational AI company that was acquired in 2019. He has worked with language models for more than a decade and has taught everywhere from middle school classrooms to prison systems in Maryland.
