Professional
Kirill
AI Engineer specializing in LLM Fine-Tuning & LoRA | PyTorch | Unsloth | CUDA Optimization
- Based in
- Singapore
- Speaks
- 4 languages
- Skills
- 6
About
AI Engineer with a proven track record of building custom fine-tuning pipelines from scratch.
I specialize in training LoRA models on resource-constrained hardware (RTX 4060 8GB), solving complex dependency issues, and delivering production-ready GGUF models for local deployment.
What I do
- Large Language Models (LLMs)
- Fine-Tuning
- LoRA / QLoRA
- PyTorch
- Python
- CUDA
Experience
- 2026-03-06 – 2026-03-01Engineered a custom fine-tuning pipeline using Unsloth and PyTorch to train LoRA models on NVIDIA RTX 4060 (8GB VRAM), successfully adapting 3B-parameter LLMs for niche content generation.
Solved complex dependency and environment conflicts (CUDA, xFormers, bitsandbytes) to create a stable, reproducible Windows-based training workflow—a task widely considered non-trivial in the community.
Developed a complete MLOps workflow for model deployment, including conversion of Hugging Face models to GGUF format and integration with LM Studio, resulting in portable, production-ready model files.
Executed a successful pilot project, training a model on a minimal dataset (6 text examples) to replicate a client's writing style, generating a fully functional 6.4GB GGUF model.
Moscow
Education
- 2005-09-01 – 2010-08-31Lomonosov Moscow State University
Fundamental Informatics and Information Technologies
Moscow
Let's work together
Kirill · AI Engineer specializing in LLM Fine-Tuning & LoRA | PyTorch | Unsloth | CUDA Optimization
Readable by AI assistants: view Kirill's card