Photo by the author
# Entry
If you’ve used LLM for a variety of tasks, you’ve probably noticed that the answer often depends on how you write the prompt. This is what we call rapid engineering. The way you provide instructions can be the difference between a vague answer and a precise, actionable answer. I know that rapid engineering can sometimes seem tough. It’s not just pure science; it’s a combination of science and art, which means you have to experiment to see what works best in each situation. Don’t worry, I’ll cover you in this article.
We’ll get through 7 proven recipes which you can bookmark and utilize for your own tasks. I won’t cover every domain here, but I will focus on 7 different areas. If any of these closely align with what you’re working on, try it out and let me know how it goes in the comments. Here we go.
# 1. Job Application & Career → Person + Personalization Hint
Generic cover letters are quite uncomplicated to spot. While I personally believe that a letter you write will read more naturally and receive a better response from the employer, I understand that this is one of the most common utilize cases. In such a scenario, add a personal touch and keep the tone natural. If you just paste your resume, it often highlights everything – even things that aren’t really vital. If you want, you can also add some key points in the structure section. Don’t just ask: “Write a cover letter for the position of ML engineer at XYZ company.” You don’t want to give the impression that your letter is identical to every other candidate’s.
Template:
You are my professional assistant. Prepare a tailored cover letter for the position [Job Title] On [Company].Details about me: [paste key skills, most relevant achievements, and work experience].
Guidelines:
– Maintain your tone: professional, confident, and at the same time natural – not too enthusiastic.
– Summarize the experience in a way that emphasizes transferable value and impact, rather than a task-by-task list.
– Structure:
1) A low introduction showing genuine interest in the role/company.
2) A concise paragraph linking my knowledge to the role requirements.
3) End the paragraph with a firm but respectful call to action.
– Keep the letter on one side.
# 2. Mathematics and logical reasoning → Chain of thought + Role + Multi-shot prompting
Most people in the community may already know what chaining of thought and multi-shot prompting is, but since many students and non-technical users utilize LLM for this purpose, I wanted to mention it explicitly. LLMs often struggle with maths if you ask them directly. For example, try asking LLM to count the number of “r”s in the word “strawberry” and you can see that it will be tough. Instead, ask for it explicitly “step by step reason” improves accuracy. Adding a few examples – worked problems – further reduces errors by ensuring a clear understanding of the reasoning process.
Template:
You are a math teacher. Solve the following problem step by step before giving your final answer.Example:
Q: If a train travels at 60 km/h for 2 hours, how far will it travel?
Ans: Step 1: Speed × time = 60 × 2 = 120 km.
Final answer: 120 kmNow solve this problem:
[Insert your math problem here]
# 3. Code Generation → Statement Layout + Constraint Hint
Coding is one of the main utilize cases of the LLM and this is why you may have heard the popular term “vibration coding”. Even experienced developers have moved to generating boilerplate code using LLM and then building on the fly. If you’ve coded before, you know that there are many ways to solve a single problem, and LLM studies sometimes make things more complicated than necessary. A bit of guidance in the form of constraints – and breaking down tasks into clear inputs, outputs and requirements – will make the results actionable.
Template:
You are a senior software engineer. Write Python code to perform the following task using {constraint}.Task: {describe what the code should do}
Requirements:
Input format: {specify}
Output format: {specify}
Edge cases to handle: {list them}Provide only spotless, commented code.
# 4. Learning and tutoring → Socratic method + Guided teaching
Many people utilize LLMs as a learning tool because of the flexibility they provide and the way they can be easily adapted to suit their preferred structure. Different teaching methods work differently on people, but one approach that I find useful and widely used in education is that learning is not just one-way. Instead, the teacher asks questions to check understanding and then explains or explains further. This makes the process interactive and prevents passive reading.
Template:
You are a patient teacher. Instead of giving me the answer directly, walk me step by step using questions I can answer. Then, based on my answers, explain the solution clearly.Subject: {Insert subject}
Start teaching:
# 5. Inventive Writing and Storytelling → Controlled creativity with Persona + Style
One of the main utilize cases that emerged for LLM was the augment in children’s content due to its ability to generate engaging stories. You may also have noticed that AI-powered videos on YouTube are following the same trend. Creating a story is pretty chilly, but if you let the model run on its own, things can easily get lost. To keep your video engaging and structured, you’ll want to set boundaries such as perspective, theme, character, and even the ending. In practice, this works much better for inventive tasks.
Template:
You are a gifted storyteller. Write a low story (approximately 400 words) in the style of magical realism.Perspective: first person
Theme: discovering the hidden world in the ordinary
Audience/level of complexity: children (uncomplicated)
Ending: End with a surprising twist.
# 6. Brainstorming and idea generation → Divergent and convergent thinking
When it comes to creativity, one of the most effective ways to utilize the LLM is to brainstorm. However, if you only ask for “ideas”, the model may throw out a random list that is either too broad or impractical. A better way is to utilize the same process used in real brainstorming sessions: first broaden the topic and generate as many raw ideas as possible (divergent thinking), then narrow and refine the best ones into viable solutions (convergent thinking). This way you will get both creativity and structure in the result.
Template:
Step 1: Generate 10 raw, unfiltered ideas for [topic].
Step 2: Choose the 3 most practical ideas and develop each of them into a detailed plan.
# 7. Business and Strategy → Structured consultant-style prompt
Many people also utilize LLM for business-related tasks, whether it is market research, planning or strategy building. The problem is that if you ask a vague question, e.g “How to improve your business?” you will usually get a generic answer that won’t really support. A way to get more actionable and clear results is to formulate your prompts in a structured format, similar to how consulting firms present their analyses. This keeps the answer focused, avoids unnecessary confusion, and makes it actionable.
Template:
You are a strategy consultant. Provide a structured, three-part analysis for [business challenge].Current situation: key facts, market context or available data
Key Challenges: The main problems or obstacles to solve
Recommended strategy: 3 actionable steps that can be implemented directly
Kanwal Mehreen is a machine learning engineer and technical writer with a deep passion for data science and the intersection of artificial intelligence and medicine. She is co-author of the e-book “Maximizing Productivity with ChatGPT”. As a 2022 Google Generation Scholar for APAC, she promotes diversity and academic excellence. She is also recognized as a Teradata Diversity in Tech Scholar, a Mitacs Globalink Research Scholar, and a Harvard WeCode Scholar. Kanwal is a staunch supporter of change and founded FEMCodes to empower women in STEM fields.