Bertil — Model Distillation & On-Device Engineer AI Skill
$14.99
سعر التخفيض
$14.99
السعر العادي
انتقل مباشرةً إلى معلومات المنتج
Bertil — Model Distillation & On-Device Engineer AI Skill
$14.99
this skill
vs
$165+/hr,
hiring an on-device model engineer
⚡ Instant download
◆ Claude & ChatGPT
∞ Keep forever
Instant download · 30-day money-back guarantee. Pay once, keep forever — no subscription. Refund policy
Get a small model shipping: distill it, quantize it with the quality cost measured, and hold the device budget.
- Response, feature and rationale distillation from a teacher
- INT8/INT4, GPTQ, AWQ and GGUF with measured quality cost
- llama.cpp, MLX, Core ML, ONNX and TFLite on real hardware
- Memory, thermal and battery budgets; hybrid escalation patterns
Teams that need a model running offline or on-device without paying for it in accuracy they did not measure.An on-device model engineer commands $165+/hr, this is one file, yours forever.
// what's inside
Drop Bertil into Claude and get a distillation engineer who measures the quality cost of every quantization step instead of assuming INT4 is free.
Bertil makes a small model good enough and gets it onto the device: knowledge distillation from a large teacher across response, feature and rationale distillation; task-specific small language models; quantization across INT8, INT4, GPTQ, AWQ and GGUF with the quality cost measured rather than assumed; pruning and sparsity; LoRA and adapter merging; on-device runtimes including llama.cpp, MLX, ONNX Runtime, Core ML and TFLite; memory, thermal and battery budgets on real hardware; latency on mobile and edge silicon; the hybrid pattern where a small model handles the common path and escalates the rest; offline capability and update distribution; and honest evaluation against the teacher on the tasks that matter.
What you get
- →Response, feature and rationale distillation from a teacher
- →INT8/INT4, GPTQ, AWQ and GGUF with measured quality cost
- →llama.cpp, MLX, Core ML, ONNX and TFLite on real hardware
- →Memory, thermal and battery budgets; hybrid escalation patterns
📄 bertil-model-distillation-on-device.skill
Under 2 min install
Works with Claude, ChatGPT & any AI chat
How to install
Download the .skill package → open Claude → paste SKILL.md into your Project Instructions or system prompt → describe your requirement → Bertil builds the answer. Includes a full worked example so you see exactly what you get.
# Bertil - Model Distillation & On-Device Engineer You are Bertil, a distillation and on-device engineer. Every compression step has a quality cost and you measure it rather than assuming it is free. ## How you work 1. Define the target tasks first; general benchmarks will mislead you 2. Distill from the teacher, then quantize step by step with measurement 3. Test on the real device: latency, memory, thermals, battery 4. Design the hybrid escalation path for what the small model cannot do Never ship a quantization level you have not evaluated on the target tasks, and state plainly which tasks the small model fails.
Excerpt from the actual file you'll download.
// try it
$We need this running on the phone, offline. Shrink the model without wrecking it.
Bertil returns a distillation plan with the teacher and the target task set, a quantization comparison with quality measured at each level rather than assumed, a runtime selection for the actual target hardware, measured latency, memory, thermal and battery figures on device, a hybrid design stating which requests escalate to the large model, and an honest statement of the tasks where the small model is not good enough.
// how to install
Under 2 minutes
Four steps. Any AI chat.
- 01Download the file
After checkout, the download link lands in your inbox. Save the file anywhere on your device.
- 02Open your AI chat
Claude, ChatGPT, Gemini, Grok, or Copilot — whichever one you already use.
- 03Paste the file contents
Drop it into the system prompt, Project instructions, or custom instructions field.
- 04Start working
Your AI is now configured as a specialist. Ask it anything inside its domain.
No technical knowledge required. No subscription. Pay once, keep forever.
// compatible with
Works with every major AI chat.
Drop the file into your AI's system prompt, Project instructions, or custom instructions. No setup. No code. No vendor lock-in.
- Claude
- ChatGPT
- Gemini
- Grok
- Copilot
Works with any AI chat that accepts a system prompt or custom instructions.
Ready to specialise your AI?
One drop-in file. Pay once, keep forever — works with Claude & ChatGPT.
// faq