About Me

Hello! I’m Osman, the developer behind Candid Garden.

I’m based in Munich, Germany. I earned my Bachelor’s degree in Economics from Yildiz Technical University in Istanbul, Turkey, before moving to Munich to pursue a Master’s in Economics at Ludwig-Maximilians-Universität München (LMU).

After completing my master’s, I was eager to bridge my interest in art with digital humanities. I applied to the PhD program at LMU—and was accepted. My research focused on automating the tagging of art images, building on a large dataset of crowd-sourced tags from the Artigo project, which was led by my professor and served as the foundation for this work.

My hypothesis was that a machine learning model could be trained not only to detect pre-iconographic elements—like chair, clock, or human—in representational paintings, but also to recognize more abstract concepts. I imagined making it possible to ask questions like:

and have the system return not only obvious portrayals but also subtle, symbolic, and hidden ones.

I wanted to build a tool that democratizes the question:

“What do you see in this painting?”

A tool that helps anyone—not just art historians—see and understand the rich visual language and symbolism in art, without needing to navigate academic jargon or cultural elitism.

Academically, the project aimed to support art historians in a very real problem: the sheer volume of contemporary visual production. More art is being created today in a single month than what's housed in museums worldwide. No human can keep up. If researchers want to write about specific themes, they need a way to filter and surface relevant works from vast collections. I envisioned a model that could learn a new concept from just a few hundred examples—curated by the researcher—and then apply that understanding across large datasets. This was the question that got me into the PhD program in 2018, at the dawn of the AI revolution.

At the time, I was experimenting with pretrained models and already seeing promising results.

But while all this exciting work was unfolding, life behind the scenes was deeply unstable. My PhD wasn’t funded, and I faced visa issues. I worked odd jobs, moved every few months—sleeping on friends’ couches, staying in temporary sublets—because I couldn’t afford long-term housing. Still, I pushed through, fueled by passion and the belief that this situation was temporary.

After about a year and a half, I hit a wall.

I hesitate to call it burnout, because burnout is often about doing something relentlessly without hope. For me, it felt more like a slow erosion—of confidence, motivation, and joy. I became withdrawn, cynical, and emotionally exhausted. Then the pandemic hit. I needed stability. A steady income. A long pause.

I’m incredibly grateful to my friends who helped me during that time. I love you all.

I found a job. Fast-forward four and a half years: I’m no longer a PhD student, and the world has changed in ways I couldn’t have imagined. CLIP model from OPENAI, SEGNET Model from Meta and of course Large Language Models (LLMs) have transformed what's possible. What I once dreamed of building during my PhD now exists—more powerful than I could have ever achieved back then. It’s incredible. And it reignited my motivation.

That’s how Candid Garden was born.

I’m developing this project using today’s state-of-the-art AI technologies, hoping to create something that feels both deeply personal and globally useful. I want to help people see more in art—to compare their perspectives with others, connect with the emotions and history embedded in each work, and begin a journey of reflection and discovery.

I dream of building end-to-end open-source AI models made by people, for people—tools rooted in empathy, creativity, and shared understanding.

And sometimes, I imagine myself as a librarian in Borges’ infinite Library… or a historian wandering Ursula K. Le Guin’s Hainish worlds, where memory, stories, and possibility are sacred.

Thank you for visiting.

If this project speaks to you, I’d love to hear from you:

📬 hey[at]candidgarden.com