I will finetune, quantize, and distill llms and vlms

Your AI model costs too much to run, or doesn't behave consistently enough. Fine-tuning, quantization, and distillation fix both for language models and vision-language models alike. I'll tell you honestly which one your problem needs.

What you get:

• Fine-tuning only where it's the right tool if prompting or RAG solves it cheaper, I'll say so first
• LoRA/QLoRA fine-tuning on open-weight LLMs (Llama, Qwen, Mistral) and VLMs (Qwen-VL, LLaVA) most jobs run on a single GPU for a few dollars in compute
• Quantization (4-bit/8-bit) to cut inference cost and latency with minimal quality loss
• Knowledge distillation train a smaller, cheaper model to match your expensive frontier-model calls at scale
• Evaluation included: task-specific metrics, not just training loss

Background: production ML engineering this is a cost/quality optimization problem, not a checkbox.

Send me your current setup and cost pain point, and I'll tell you what actually moves the needle.