Descript India Prompt Engineer Interview: Rounds, Questions and Preparation

What the job involves

Prompt engineers design, test, and refine instructions that make large language models produce reliable, useful outputs for products like AI-powered video editing, transcription, and content generation. Day to day, they write and iterate on prompts, run evaluations, analyze failure cases, and work with product and engineering teams to ship AI features. In a deeptech sector, they also stay current with model capabilities and safety considerations.

Role and technical knowledge

1. 1

   What is prompt engineering, and how does it differ from traditional software engineering?

   Explain that prompt engineering focuses on designing natural language instructions and examples to guide LLM behavior, while software engineering builds deterministic systems. Mention iteration and evaluation as core differences.

2. 2

   How would you design a prompt to summarize a long video transcript into key points for a video editing tool?

   Cover role instruction, output format, length constraints, and handling of timestamps. Mention testing with varied transcripts and using few-shot examples for consistency.

3. 3

   What is chain-of-thought prompting, and when would you use it?

   Describe how it encourages step-by-step reasoning for complex tasks like math or multi-step logic. Note that it may not be needed for simple extraction tasks and can increase token usage.

4. 4

   How do you evaluate whether a prompt is working well?

   List metrics like accuracy, relevance, fluency, and safety. Mention creating a test set, human review, and automated scoring with tools like exact match or LLM-as-judge.

Practical scenarios

1. 5

   A prompt that generates video titles starts producing offensive outputs. How do you fix it?

   Suggest adding explicit safety instructions, negative examples, and a post-filter. Mention testing with adversarial inputs and iterating with a diverse evaluation set.

2. 6

   You need to reduce hallucinations in a Q&A feature for video content. What steps do you take?

   Propose retrieval-augmented generation using the video transcript as context, instructing the model to say 'I don't know' when unsure, and adding citations. Test with questions outside the transcript.

3. 7

   How would you adapt a prompt for a smaller, faster model versus a larger, more capable one?

   Explain that smaller models may need simpler instructions, more examples, and stricter formatting. Larger models can handle nuanced reasoning but may need cost and latency trade-offs.

4. 8

   Describe how you would use few-shot examples to improve a prompt for generating video descriptions.

   Cover selecting diverse, high-quality examples, ordering them by relevance, and keeping them concise. Mention testing with and without examples to measure improvement.

Behavioural and collaboration

1. 9

   Tell me about a time you had to explain a technical concept to a non-technical person.

   Use the STAR method. Highlight simplifying jargon, using analogies, and checking understanding. Relate it to working with designers or product managers.

2. 10

   How do you handle feedback that your prompt is not performing as expected?

   Show openness to feedback, a systematic debugging approach, and willingness to iterate. Mention asking clarifying questions and documenting changes.

3. 11

   Describe a situation where you had to meet a tight deadline on a project.

   Focus on prioritization, communication, and delivering a minimum viable solution. Mention learning from the experience for future planning.

4. 12

   Why do you want to work in AI and deeptech, specifically in a role like prompt engineering?

   Connect personal interest in language and AI, desire to shape product experiences, and eagerness to learn in a fast-evolving field. Avoid generic answers; mention specific aspects of the sector.

HR and company fit

1. 13

   What do you know about our company and how does this role fit into our mission?

   Research the company's products and AI use cases. Explain how prompt engineering contributes to user experience and product differentiation.

2. 14

   How do you stay updated with the latest developments in LLMs and prompt engineering?

   Mention following AI blogs, research papers, forums, and experimenting with new models. Show curiosity and hands-on learning.

3. 15

   Where do you see yourself in three years in this field?

   Express a desire to deepen expertise in prompt design and evaluation, possibly leading projects or mentoring. Align growth with company needs.

4. 16

   What are your salary expectations for this role?

   Research industry standards for freshers in AI roles. Give a range and emphasize flexibility based on overall compensation and learning opportunities.