Hi, I’m Yelaman. Or just Yela.
I’m an AI and software engineer based in Sydney. I work on applied AI at StarRez and recently submitted my PhD thesis in Computer Science.
I’m interested in machine learning research and engineering, particularly reasoning, generative and multimodal models, reinforcement learning, and model training.
About
I started in software engineering and have worked across backend systems, NLP, distributed systems, and engineering leadership. Over time, my work moved closer to machine learning and AI research.
I like problems where research ideas meet real systems: experiments, data, evaluation, training, tooling, and production constraints.
Work
StarRez
I work as a Senior AI Engineer, building LLM-powered product features, multimodal applications, and evaluation systems for generative AI.
Earlier engineering work
Before StarRez, I spent several years in software engineering roles across trading systems, distributed backend systems, NLP, and engineering leadership.
Research
PhD research
My PhD work studies information elicitation in optimisation-modelling dialogues, including synthetic data, question-asking strategies, reinforcement learning, and multimodal representations.
Reasoning and model training
I experiment with ARC-AGI, mathematical reasoning, supervised fine-tuning, reinforcement-learning-based post-training, and model evaluation using open models and training stacks.
Generative models
I’ve also been exploring flow matching, masked diffusion, diffusion language models, and other model architectures through independent experiments and implementations.
Qurt
Qurt is a model-agnostic desktop AI coworker I built to explore model behaviour, tools, interfaces, and workflows.
Community
I organize seminars and reading sessions in the DSML.KZ machine-learning community, usually around recent research and model architectures.
A couple of talks I’ve given at DSML.KZ seminars: