AI & Machine Learning Engineer
Sona
Mansuryan
Building intelligent systems by combining machine learning, research-driven thinking, and real-world problem solving.
Scroll
Building intelligent systems by combining machine learning, research-driven thinking, and real-world problem solving.
I'm a researcher and developer focused on the intersection of machine learning theory and practical applications. My work spans generative models, reinforcement learning, and NLP fields where algorithms begin to exhibit truly fascinating emergent behavior.
From implementing classic algorithm theory to training CycleGANs that transform photographs into Van Gogh paintings, I pursue projects that push my understanding of what neural networks can accomplish.
Currently deepening my expertise in generative AI and exploring how large language models reshape our interaction with software.
An AI-powered customer support agent with Armenian language voice capabilities — bridging speech recognition and NLP.
An implementation of CycleGAN for unpaired image-to-image translation — transforming ordinary photos into Van Gogh paintings.
Training an AI agent to master the classic Chrome Dinosaur game using the NEAT genetic algorithm.
A collection of RL algorithms (Q-Learning, Policy Gradients) and environments implemented from scratch.
ML-powered text classification system utilizing advanced NLP to identify misinformation and analyze credibility.
Deep learning CNN model designed to detect and classify crop diseases from leaf imagery for precision agriculture.
A generation engine exploring modern diffusion architectures to convert text prompts into detailed visuals.