Software Engineer at Soft Strategy (Java / Spring Boot). I build at the intersection of backend systems and applied AI — fine-tuning open-source LLMs and taking ML from research to production.
- Software Engineer @ Soft Strategy — enterprise backends in Java / Spring Boot (railway sector).
- Growing toward AI Engineering — LLM fine-tuning, evaluation, agents.
Fine-tuned Qwen3.6-27B (DoRA) to write COBOL, on a pipeline where every example is compiled and executed against real GnuCOBOL before entering the dataset. 88.4% compile (beats published open-source SOTA, 73.95%) · 36.3% Pass@1 (2× GPT-4o) · <$120 GPU cost → Write-up · Model on Hugging Face
Thyroid-Nodule-AI — Bachelor's thesis
Two-stage deep-learning CAD pipeline for thyroid nodule diagnosis (7,000+ ultrasound images). A self-supervised Vision Transformer (DINOv3) outperforms CNN baselines. AUC 0.932 · Sensitivity 94.70% · Grad-CAM explainability · K-TIRADS integration · GUI demo → Read the thesis
Autonomous AI agent for Linux: LLM reasoning core + n8n orchestration, RAG memory, Human-in-the-Loop via Telegram, React dashboard.
Languages
AI & Data Science
Backend & Infrastructure
