I am an aspiring researcher in the field of Machine Learning and Deep Learning (ML/DL).
My current research focuses on LLM-driven evolution and methods to optimize its performance while reducing computational costs. I am also interested in Recursive Self-Improvement and novel AI architectures. Previously, my work focused on evolutionary neural architecture search (NAS) applied to computer vision tasks.
- 🧬 LLM-Driven Evolution & Self-Improvement: Enhancing LLM-driven evolution through persistent memory architectures and adaptive curriculum learning.
- 📉 Efficiency & Optimization: Accelerating search runtimes, reducing compute overhead, and optimizing resource allocation in evolutionary pipelines.
- 🔄 Recursive Self-Improvement: Investigating self-correcting systems capable of iterative, autonomous meta-learning.
- Core ML/DL: PyTorch, HuggingFace Transformers, scikit-learn, Optuna
- LLM & Multi-Agent: LangGraph, Langchain, vLLM
- Data Science & Viz: NumPy, Pandas, Matplotlib, Seaborn
- Backend & DBs: FastAPI, Redis, PostgreSQL, ChromaDB
- DevOps & Tools: Git, Docker, Docker-compose
-
🧠 EvoMem (GigaEvo Core) – Main Contributor Developed a general-purpose memory system designed for the GigaEvo framework, focusing on robust context management and persistence for evolutionary ML processes.
-
⚔️ HeroBench – AI Agent Developer A comprehensive benchmark evaluating long-horizon planning and structured reasoning capabilities of LLM agents within virtual worlds.
-
📄 LLM Trust Simulation – Independent Researcher A simplified reproduction project measuring economic trust behaviors in small, local LLMs utilizing the Dictator Game paradigm. Developed during the "Summer with AIRI 2025" research school.
💬 Feel free to explore my research website above for additional ways to connect or collaborate!