Perplexity AI India Data Scientist Interview: Rounds, Questions and Preparation

What the job involves

Data Scientists in AI and deeptech build and evaluate models that power search, recommendations, and generative AI features. At companies like Perplexity AI India, they work on retrieval, ranking, and large language model applications, turning user queries and feedback into measurable product improvements. Day to day involves data pipelines, experimentation, model fine-tuning, and cross-functional collaboration with engineers and product managers.

Role and technical knowledge

1. 1

   How would you design a retrieval system for a search engine that answers natural language questions?

   Describe query understanding, document indexing, embedding-based retrieval, and re-ranking. Mention evaluation metrics like recall@k and MRR.

2. 2

   Explain the difference between precision and recall in the context of a recommendation system. When would you prioritize one over the other?

   Define both metrics and give examples. Discuss trade-offs based on user experience goals, such as avoiding irrelevant recommendations versus missing relevant ones.

3. 3

   What are embeddings, and how are they used in modern NLP applications like semantic search?

   Explain dense vector representations, training objectives like contrastive learning, and how cosine similarity enables semantic matching beyond keyword search.

4. 4

   How does a transformer model handle long documents differently from a recurrent neural network?

   Highlight self-attention, parallel processing, and positional encodings. Discuss limitations like quadratic complexity and techniques like chunking or sparse attention.

Practical scenarios and problem solving

1. 5

   You have a dataset of user queries and click logs. How would you build a model to rank search results?

   Outline feature engineering from clicks, training a learning-to-rank model, and offline evaluation. Mention handling position bias and using counterfactual methods.

2. 6

   Your A/B test shows a 2% increase in click-through rate but a 5% drop in user satisfaction scores. How do you interpret and act on this?

   Discuss metric trade-offs, segment analysis, and qualitative feedback. Suggest running a longer test or investigating whether clicks are misleading.

3. 7

   How would you detect and mitigate bias in a language model used for generating answers?

   Cover data auditing, bias metrics, prompt engineering, and fine-tuning with balanced data. Mention human evaluation and red-teaming.

4. 8

   You need to deploy a model that answers questions about recent news. How do you ensure it stays up to date?

   Propose a retrieval-augmented generation pipeline with frequent index updates. Discuss caching, freshness metrics, and fallback strategies.

Behavioural and teamwork

1. 9

   Describe a time you had to explain a complex technical concept to a non-technical person. How did you ensure they understood?

   Use a specific example, emphasize analogies and visual aids, and show how you checked for understanding and adapted your explanation.

2. 10

   Tell me about a project where you had to learn a new tool or technique quickly. How did you approach it?

   Highlight self-learning strategies, such as online courses, documentation, or peer mentoring. Show initiative and a concrete outcome.

3. 11

   Give an example of a time you received critical feedback on your work. How did you respond?

   Show openness to feedback, a growth mindset, and the steps you took to improve. Avoid being defensive.

4. 12

   How do you prioritize tasks when working on multiple data science projects with tight deadlines?

   Discuss impact versus effort, communication with stakeholders, and using tools like kanban boards. Mention flexibility and regular check-ins.

HR and company fit

1. 13

   Why do you want to work in AI and deeptech, and why at a company like Perplexity AI India?

   Connect personal interest in AI advancements to the company's mission. Mention specific products or technologies and how your skills align.

2. 14

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

   Show ambition to grow technically and take on more responsibility, while staying grounded in learning and contributing to team goals.

3. 15

   What do you know about our company's approach to search and answer generation?

   Demonstrate research on the company's use of retrieval-augmented generation, citation, and conversational search. Relate it to your own projects.

4. 16

   How do you handle working in a fast-paced environment with changing priorities?

   Emphasize adaptability, communication, and focusing on high-impact tasks. Give an example from internships or college projects.