I use deep learning frameworks to build original and customized machine learning engines, including object-detection models(CNN), time-based models (RNN), and Bayesian optimization.
I am actively training Large Language Models including Qwen-3, Phi-4, and Llama-3. I primarily source datasets from Hugging Face and evaluate performance using benchmarks such as SuperGPQA, HLE, Agentbench, and StructEval.
I also train CNN-based models like YOLO. I mainly reference datasets from Roboflow and have worked on practical applications including bacterial detection in medical domains and automated danger prediction systems for kindergartens that detect adults, children, and objects.
I utilize data analytics APIs including Google Analytics, and frameworks like LangChain, DSPy, and Mastra to build AI applications. I have created analytics systems to increase blog user acquisition and AI agents for recommending promising talent.
I'm interested in modern web development with TypeScript and the React ecosystem for building scalable user interfaces, including testing frameworks. This knowledge will be valuable for my goal of becoming a Product Manager.
I'm interested in modern backend development with Go, Rust, and Python. Recently, I'm studying architectural strategies to make application readable (especially for Python projects) and maintainable.
To become a strong engineer, I believe I must study how to operate and scale applications and ML models.
As long as data is properly organized, you can create reasonably good models even with suboptimal approaches. I believe that handling massive datasets is unavoidable for creating better models.