EXL Analytics NLP Engineer Interview: Rounds, Questions and Preparation

What the job involves

An NLP Engineer in AI and deeptech builds and deploys models that process and understand human language, such as text classification, named entity recognition, sentiment analysis, and question answering. At companies like EXL Analytics, you would work on client-facing projects in domains like insurance, healthcare, and banking, turning unstructured text into actionable insights. Day to day involves data cleaning, model training, evaluation, and integrating NLP pipelines into production systems.

Role and Technical Knowledge

1. 1

   Explain how a transformer model works and why it is preferred over RNNs for NLP tasks.

   Cover self-attention, parallel processing, and handling long-range dependencies. Mention advantages like scalability and contextual embeddings.

2. 2

   What is the difference between stemming and lemmatization? When would you use each?

   Stemming cuts words to root form crudely, lemmatization uses vocabulary and morphology. Use lemmatization for semantic tasks, stemming for speed in search.

3. 3

   How would you handle out-of-vocabulary words in a text classification model?

   Discuss subword tokenization (BPE, WordPiece), using pre-trained embeddings, and fallback to unknown token with proper handling.

4. 4

   Describe the architecture of BERT and how it is fine-tuned for a downstream task like sentiment analysis.

   Mention bidirectional encoder, masked language modeling, and adding a classification head. Explain fine-tuning with task-specific data.

Practical Scenarios

1. 5

   You have a dataset of customer complaints with imbalanced classes. How would you build an NLP model to classify them?

   Discuss resampling, class weights, using F1-score, and trying models like logistic regression or fine-tuning BERT. Mention data augmentation if needed.

2. 6

   A client wants to extract names of drugs and dosages from medical transcripts. How would you approach this NER problem?

   Talk about annotating data, using spaCy or Hugging Face for custom NER, handling domain-specific vocabulary, and evaluating with entity-level F1.

3. 7

   Your text summarization model produces fluent but factually incorrect summaries. How would you debug and improve it?

   Check for hallucination, use abstractive vs extractive methods, add factual consistency metrics, and consider reinforcement learning from human feedback.

4. 8

   How would you deploy an NLP model as a REST API for real-time predictions?

   Mention Flask or FastAPI, containerization with Docker, handling input preprocessing, and monitoring for latency and drift.

Behavioural and Project Questions

1. 9

   Tell me about an NLP project you built. What was your role and what challenges did you face?

   Use STAR format. Highlight data collection, model choice, evaluation, and how you overcame a specific issue like low accuracy or limited data.

2. 10

   Describe a time you had to learn a new NLP technique quickly for a deadline.

   Show initiative and learning ability. Mention resources used, how you applied it, and the outcome.

3. 11

   How do you ensure your NLP models are fair and unbiased?

   Discuss checking for bias in training data, using fairness metrics, and techniques like reweighting or adversarial debiasing.

4. 12

   Why do you want to work as an NLP Engineer in a consulting or analytics firm like EXL?

   Connect interest in solving real-world business problems with NLP, mention exposure to diverse domains, and show enthusiasm for client impact.

HR and Company Fit

1. 13

   Where do you see yourself in three years as an NLP Engineer?

   Show growth mindset: from executing tasks to leading small projects, deepening expertise in transformers or domain-specific NLP.

2. 14

   How do you handle tight deadlines and multiple client requests?

   Prioritize using impact and urgency, communicate transparently, and use tools like Jira or Trello. Give a brief example.

3. 15

   Are you comfortable working with clients from different time zones?

   Be positive and flexible. Mention willingness to adjust schedules and communicate clearly asynchronously.

4. 16

   What do you know about EXL Analytics and why does it interest you?

   Research the company's focus on data-driven solutions in insurance, banking, and healthcare. Mention its use of AI and NLP for client outcomes.