20 production recipes · copy, paste, ship

Fine-tune. Serve. Own your weights.

End-to-end guides to fine-tune a private model and serve it per-token on shared GPUs. Each takes minutes and costs cents — your weights stay yours.

Customer supportLoRA

Fine-tune a support-ticket classifier

Route and tag customer support tickets with a private Llama-3.1-8B LoRA — trained in minutes, served for cents.

Llama 3.1 8B Instruct~12 min~$0.30 / M tokens
Open recipe
Developer toolsQLoRA

Build a private SQL copilot

Turn natural language into your schema's SQL with a Qwen3-8B QLoRA — your queries never leave your tenant.

Qwen3 8B~15 min~$0.30 / M tokens
Open recipe
MarketingLoRA

Train a model in your brand voice

Fine-tune a private writer that sounds exactly like your brand — on your own copy, never shared.

Mistral Small 3.1 24B~20 min~$0.45 / M tokens
Open recipe
Developer toolsLoRA

Fine-tune a code-review assistant

A private reviewer trained on your codebase's conventions — flags issues in your house style, in your tenant.

Qwen3 8B~14 min~$0.30 / M tokens
Open recipe
Knowledge workLoRA

Train a domain document summarizer

Summarize contracts, papers, or reports in your domain's style with a private Llama-3.1-8B adapter.

Llama 3.1 8B Instruct~12 min~$0.30 / M tokens
Open recipe
Conversational AIFast LoRA

Fast intent classifier on Gemma 3 12B

Mint an intent classifier in minutes with fast fine-tuning on Gemma 3 12B.

Gemma 3 12B IT~4 min~$0.35 / M tokens
Open recipe
Retrieval / RAGLoRA

Fine-tune a RAG reranker

Reorder retrieved passages by true relevance to your domain — a private Qwen3-8B LoRA that lifts answer quality without touching your retriever.

Qwen3 8B~14 min~$0.30 / M tokens
Open recipe
Agents / toolsLoRA

Train a reliable function-calling model

Teach a private Llama-3.1-8B to emit valid tool calls for YOUR API surface — fewer malformed calls, no schema drift.

Llama 3.1 8B Instruct~16 min~$0.30 / M tokens
Open recipe
Customer supportLoRA

Fine-tune an email triage + responder

Classify inbound email and draft on-policy replies with a private Mistral-Small adapter — your inbox never leaves your tenant.

Mistral Small 3.1 24B~22 min~$0.45 / M tokens
Open recipe
Legal / knowledge workLoRA

Extract contract clauses to JSON

Pull parties, terms, dates and obligations out of contracts into structured JSON — a private Llama-3.1-8B LoRA for your paper.

Llama 3.1 8B Instruct~15 min~$0.30 / M tokens
Open recipe
E-commerceFast LoRA

Fast e-commerce product categorizer

Map messy product titles to your taxonomy in minutes with fast fine-tuning on Gemma 3 12B.

Gemma 3 12B IT~5 min~$0.35 / M tokens
Open recipe
AnalyticsLoRA

Aspect-based sentiment analysis

Score sentiment per aspect (price, quality, service…) from reviews with a private Llama-3.1-8B LoRA.

Llama 3.1 8B Instruct~12 min~$0.30 / M tokens
Open recipe
Conversational AILoRA

Build a custom-persona chatbot

A chat model that stays in character and on-policy — fine-tune a private Mistral-Small persona on your transcripts.

Mistral Small 3.1 24B~21 min~$0.45 / M tokens
Open recipe
Developer toolsLoRA

Generate docstrings in your style

Auto-write docstrings that match your repo's conventions with a private Qwen3-8B LoRA.

Qwen3 8B~14 min~$0.30 / M tokens
Open recipe
Information extractionLoRA

Custom named-entity extractor

Extract your domain's entities (SKUs, gene names, ticket IDs…) with a private Llama-3.1-8B LoRA — no off-the-shelf NER fits your schema.

Llama 3.1 8B Instruct~13 min~$0.30 / M tokens
Open recipe
LocalizationLoRA

Domain-specialized translation

Translate in your domain's terminology and register with a private Qwen3-8B LoRA — generic MT gets your jargon wrong.

Qwen3 8B~16 min~$0.30 / M tokens
Open recipe
ProductivityLoRA

Meeting transcript → action items

Turn raw transcripts into owners, decisions and action items with a private Llama-3.1-8B LoRA.

Llama 3.1 8B Instruct~14 min~$0.30 / M tokens
Open recipe
Trust & safetyFast LoRA

Fast content moderation classifier

Flag policy violations against YOUR rules in minutes with fast fine-tuning on Gemma 3 12B.

Gemma 3 12B IT~5 min~$0.35 / M tokens
Open recipe
HR techLoRA

Parse resumes to structured JSON

Extract skills, roles and dates from resumes into clean JSON with a private Qwen3-8B LoRA.

Qwen3 8B~13 min~$0.30 / M tokens
Open recipe
Model alignmentDPO

Align a model with DPO preference data

Make your assistant prefer your good answers over your bad ones with Direct Preference Optimization on Llama-3.1-8B.

Llama 3.1 8B Instruct~25 min~$0.30 / M tokens
Open recipe