Prompt Engineering Techniques: LLM Optimization Guide

Prompt engineering is the art and science of systematically designing inputs to get the desired output from AI models. Correct prompt strategies can increase model performance by up to 50%.

Prompt Engineering Fundamentals

Prompt Anatomy

An effective prompt consists of the following components:

1. System Instruction: Model's role and behavioral rules
2. Context: Background information related to the task
3. Example(s): Desired output format examples
4. Task: Specific request or question
5. Output Format: Expected response structure

Basic Prompt Structure

1[Role Definition]
2You are a {field of expertise} expert.
3
4[Context]
5{Relevant information}
6
7[Task]
8{Specific request}
9
10[Format]
11{Expected output format}

Basic Prompt Techniques

1. Zero-Shot Prompting

Direct task definition without giving examples:

Summarize the following text:
{text}

Use cases:

• Simple tasks
• General knowledge questions
• Classification

2. Few-Shot Prompting

Task definition with examples:

1Text: "This product is great, I am very satisfied"
2Sentiment: Positive
3
4Text: "It was a terrible experience, I do not recommend it"
5Sentiment: Negative
6
7Text: "It's okay for the price"
8Sentiment: ?

Best Practices:

• 3-5 examples are usually sufficient
• Add various edge cases
• Order examples randomly

3. Chain-of-Thought (CoT)

Step-by-step thinking:

1Question: A store has 15 apples. 8 apples were sold, 
2then 6 more apples arrived. How many apples are there?
3
4Let's think step by step:
51. Initially there are 15 apples
62. 8 apples sold: 15 - 8 = 7 apples left
73. 6 apples arrived: 7 + 6 = 13 apples
8
9Answer: 13 apples

4. Self-Consistency

Different reasoning paths for the same question:

1Solve this problem in 3 different ways and choose the most consistent answer:
2
3[Problem]

5. Tree of Thoughts (ToT)

Branched thought tree:

1Problem: {complex problem}
2
3Thought 1: {approach A}
4  → Sub-thought 1.1: {detail}
5  → Sub-thought 1.2: {detail}
6  
7Thought 2: {approach B}
8  → Sub-thought 2.1: {detail}
9  
10Evaluate and select the most suitable path.

Advanced Techniques

ReAct (Reasoning + Acting)

Thinking and action loop:

1Question: How many times larger is Istanbul's population than Paris?
2
3Thought: I need to find the population of both cities
4Action: [search] Istanbul population
5Observation: Istanbul population ~16 million
6
7Thought: Now I need to find Paris population
8Action: [search] Paris population  
9Observation: Paris population ~2.2 million
10
11Thought: I can calculate the ratio
12Action: [calculate] 16 / 2.2
13Observation: 7.27
14
15Answer: Istanbul's population is approximately 7.3 times larger than Paris.

Constitutional AI Prompting

Defining ethical and safety rules:

1System: You are a helpful assistant.
2
3Rules:
41. Do not generate harmful content
52. Do not share personal information
63. Do not help with illegal activities
74. Always be honest
8
9User question: {question}

Role Prompting

Defining specific area of expertise:

1You are a cybersecurity expert with 20 years of experience.
2You have worked as a CISO in Fortune 500 companies.
3You can explain technical details clearly and understandably.
4
5User's question: {question}

Prompt Optimization Strategies

1. Increasing Specificity

❌ Bad:

Write a blog post

✅ Good:

1Target audience: Software developers
2Topic: Docker container security
3Length: 1500-2000 words
4Tone: Technical but accessible
5Format: Introduction, 5 main sections, conclusion

2. Determining Output Format

1Provide your response in this JSON format:
2{
3  "summary": "string",
4  "key_points": ["string"],
5  "next_steps": ["string"],
6  "confidence_score": number
7}

3. Negative Prompting

Specifying unwanted behaviors:

1Do NOT do the following:
2- Give speculative information
3- Make claims without citing sources
4- Lead the user
5- Give excessively long answers

4. Use of Delimiters

Clarifying sections:

1###CONTEXT###
2{context information}
3
4###TASK###
5{work to be done}
6
7###FORMAT###
8{output format}

Model-Specific Optimizations

For GPT

1- Use System message effectively
2- Activate JSON mode: response_format={"type": "json_object"}
3- Temperature: 0.7-1.0 for creative tasks, 0.1-0.3 for analytical

For Claude

1- Use XML tags: <context>, <task>, <format>
2- Put important information at the end in long context
3- Evaluate Thinking tags

For Gemini

1- Optimize for multimodal prompts
2- Up-to-date information with Grounding
3- Adjust Safety settings

Prompt Testing and Iteration

A/B Testing Framework

1Prompt A: {version 1}
2Prompt B: {version 2}
3
4Metrics:
5- Accuracy: %
6- Consistency: 1-5
7- Latency: ms
8- Token usage: #

Prompt Versioning

1prompt_v1.0: First version
2prompt_v1.1: Typo corrections
3prompt_v2.0: CoT added
4prompt_v2.1: Output format changed

Enterprise Prompt Management

Creating Prompt Library

1/prompts
2  /classification
3    - sentiment_analysis.json
4    - intent_detection.json
5  /generation
6    - blog_writer.json
7    - code_reviewer.json
8  /extraction
9    - entity_extraction.json
10    - data_parsing.json

Prompt Template System

1class PromptTemplate:
2    def __init__(self, template, variables):
3        self.template = template
4        self.variables = variables
5    
6    def render(self, **kwargs):
7        return self.template.format(**kwargs)
8
9# Usage
10sentiment_prompt = PromptTemplate(
11    template="Analyze sentiment: {text}",
12    variables=["text"]
13)

Common Mistakes and Solutions

Mistake 1: Excessively Vague Prompt

Problem: Model doesn't understand what you want Solution: Add specific instructions and examples

Mistake 2: Very Long Prompt

Problem: Token limit exceeded, costs increase Solution: Remove unnecessary information, use summary

Mistake 3: Conflicting Instructions

Problem: Model behaves inconsistently Solution: Prioritize and clarify rules

Mistake 4: Hallucination

Problem: Model gives made-up information Solution: Grounding, citation requirement, lower temperature

Conclusion

Prompt engineering is a critical factor in the success of AI projects. You can significantly increase model performance with correct techniques and a systematic approach.

At Veni AI, we develop customized prompt strategies for our corporate clients. Contact us for professional support.