prompt

Free

Expert system prompt for designing and deploying production LLM systems

FreeFree tier

Inputs: textOutputs: text

Type

Open Source

Visit Website GitHub (40 stars)

About prompt

LLM Architect is a detailed system prompt designed to guide a large language model to act as an expert in production LLM system design. It covers core competencies such as model selection, fine-tuning (LoRA, QLoRA, RLHF/DPO/ORPO), RAG implementation (document processing, embedding models, vector stores, retrieval optimization), production serving (quantization, KV cache optimization, speculative decoding, batching), and safety guardrails (content filtering, prompt injection defense, hallucination detection). The prompt enforces a 'start simple' principle (prompting → RAG → fine-tuning) and emphasizes measurement, data quality, cost-awareness, and safety. It includes a decision framework and critical rules for systematic escalation of complexity.

Key Features

System architecture guidance: model selection, serving infrastructure, load balancing, cost optimization

Fine-tuning expertise: LoRA/QLoRA, full fine-tuning, RLHF/DPO/ORPO, dataset preparation, hyperparameter tuning

RAG implementation: document processing, embedding models, vector stores, retrieval optimization, evaluation

Production serving: quantization (GPTQ, AWQ, GGUF), KV cache optimization, speculative decoding, batching strategies

Safety guardrails: content filtering, prompt injection defense, hallucination detection, bias mitigation, compliance checks

Decision framework: systematic escalation from prompting to RAG to fine-tuning with baseline metrics

Pros & Cons

Pros

Comprehensive coverage of LLM system design aspects from prompting to production
Emphasis on cost optimization and safety as non-negotiable
Clear decision framework to avoid unnecessary complexity
Includes practical techniques like quantization, speculative decoding, and KV cache optimization
Open source and freely available as a system prompt

Cons

Requires an LLM platform to execute; not a standalone tool
Complexity may be overwhelming for beginners without LLM deployment experience
No built-in execution environment or integration helpers

Best For

Designing production-grade LLM systems with fine-tuning and RAGOptimizing LLM serving for latency and throughput targetsImplementing safety guardrails for LLM applicationsSelecting appropriate models and infrastructure for cost and performance constraintsBuilding retrieval-augmented generation pipelines with evaluation

FAQ

What is the LLM Architect prompt?

It is a system prompt that instructs an LLM to act as an LLM Architect / Fine-tuning Specialist, covering design of production LLM systems including fine-tuning, RAG, inference serving, and safety guardrails.

What principle does the LLM Architect follow?

It follows a 'start simple' principle: prompting before RAG before fine-tuning. Complexity is escalated only when data justifies it.

What fine-tuning techniques are covered?

The prompt covers LoRA/QLoRA for parameter-efficient fine-tuning, full fine-tuning when needed, and alignment techniques like RLHF, DPO, and ORPO.

What safety measures are included?

Safety guardrails include content filtering, prompt injection defense, hallucination detection and mitigation, bias detection, and compliance checks for PII, copyright, and regulatory requirements.