Question 1

Tell me about yourself.

Accepted Answer

Use present, past, future, in about 60 to 90 seconds. Present: "I'm a Java backend developer with four years at a fintech company, where I build payment APIs." Past: "Over the last year I moved towards AI. I built a document question answering app with RAG and an evaluation set, and I used an LLM to classify support tickets at work, which cut manual triage by about 30%." Future: "Now I want to do this full time as an AI engineer, and this role fits because you build LLM features on top of real backend systems, which is exactly where my strengths meet."

Question 2

Why do you want to move into AI, and why this particular role?

Accepted Answer

Give a real reason and connect it to evidence. "In my testing work I kept seeing the same problem: thousands of test cases written and checked by hand. When I tried an LLM to draft test cases, it saved hours, but it also made mistakes I had to catch. That got me interested in building AI systems that are actually reliable. I've since completed projects in RAG and evaluation. I'm applying for this role because it focuses on shipping AI features to users, and my background in quality and testing directly helps with evaluating AI output."

Question 3

Why do you want to work here?

Accepted Answer

Show you've done homework. Mention one thing about their product, one thing about how they work, and link both to you. "I've used your app for expense reports, and I noticed the receipt scanning recently got much better, so I read your engineering blog about it. I like that your team publishes how you evaluate models, because I care about measuring AI quality. And since you serve small businesses in India, my experience building for that market would help from day one."

Question 4

You haven't had an AI job title yet. Why should we hire you over someone who has?

Accepted Answer

Don't be defensive. Turn the gap into a strength. "That's fair. What I bring is production experience: I've spent five years shipping and supporting real systems, handling incidents and working with stakeholders. Many AI projects fail at exactly that stage, after the demo. On the AI side, I've built two end to end projects with evaluation, which I can show you. So you'd get someone who learns fast and already knows how to make software work in production."

Question 5

Explain supervised, unsupervised and reinforcement learning, with an example of each.

Accepted Answer

Supervised learning learns from examples that have the right answer, called labels. For example, emails marked spam or not spam, so the model learns to predict spam. Unsupervised learning has no labels; it finds structure by itself, like grouping customers into segments by behaviour. Reinforcement learning learns by trial and error with rewards, like a game playing agent or a robot that gets a reward when it does the task well. LLMs also use reinforcement learning from human feedback to make answers more helpful.

Question 6

What is overfitting, and how do you prevent it?

Accepted Answer

Overfitting is when a model memorises the training data, including its noise, so it scores very well on training data but badly on new data. It's like a student who memorises last year's answers but can't solve a new question. You spot it when training accuracy is high and validation accuracy is much lower. To prevent it: get more data, use a simpler model, add regularisation, use cross validation, stop training early, and for neural networks use dropout.

Question 7

What is the bias variance trade off?

Accepted Answer

Bias is error from a model that's too simple, so it misses real patterns. That's underfitting. Variance is error from a model that's too sensitive to the training data, so small changes in the data change it a lot. That's overfitting. Making a model more complex usually lowers bias but raises variance. The goal is the sweet spot where total error on new data is lowest, which we find using validation data. For example, a straight line through curved data has high bias; a very deep decision tree has high variance.

Question 8

Why do we split data into training, validation and test sets?

Accepted Answer

The training set is what the model learns from. The validation set is used while building, to compare models and tune settings like depth or learning rate. The test set is kept locked away until the very end to get an honest estimate of how the model will do on new data. If you tune on the test set, you slowly overfit to it and your final number becomes too optimistic. For time based data, split by time, so you train on the past and test on the future.

Question 9

What is data leakage? Give an example.

Accepted Answer

Leakage is when information that won't be available at prediction time sneaks into training, so the model looks great in testing and then fails in production. A classic example: predicting whether a customer will cancel, and including a feature like "refund issued", which only happens after they cancel. Another is scaling features using statistics from the whole dataset before splitting, so the test set influences training. To avoid it, ask for each feature: would I know this value at the moment I make the prediction? And do all preprocessing inside the training split.

Question 10

Explain precision and recall. When would you prefer one over the other?

Accepted Answer

Precision: of everything the model flagged as positive, how many were really positive. Recall: of all the real positives, how many the model caught. For a cancer screening test or fraud detection, missing a real case is very costly, so we prefer high recall, even if we get some false alarms. For a spam filter, putting a real email into spam is annoying, so we prefer high precision. F1 score combines both when you need one number, and you can move between them by changing the decision threshold.

Question 11

Your model is 99% accurate, but the business says it's useless. How is that possible?

Accepted Answer

Probably the classes are imbalanced. If only 1% of transactions are fraud, a model that always says "not fraud" is 99% accurate and catches zero fraud. Accuracy hides that. I'd look at the confusion matrix, precision and recall for the rare class, and metrics like PR AUC. Then I'd check the business goal: maybe they need to catch 80% of fraud while keeping false alarms low enough for the review team to handle, and I'd tune the threshold for that.

Question 12

What is regularisation? What's the difference between L1 and L2?

Accepted Answer

Regularisation adds a penalty for large weights to the loss, which keeps the model simpler and reduces overfitting. L2, also called ridge, penalises the square of weights, so it shrinks all weights smoothly towards zero but rarely to exactly zero. L1, also called lasso, penalises the absolute value, which pushes some weights to exactly zero, so it also does feature selection. In neural networks we also use dropout and weight decay, which is L2 in practice.

Question 13

What's the difference between a random forest and gradient boosting?

Accepted Answer

Both combine many decision trees. A random forest trains lots of deep trees independently on random samples of rows and features, then averages their votes. That reduces variance and is hard to break. Gradient boosting trains small trees one after another, where each new tree fixes the errors of the ones before it. Boosting, like XGBoost or LightGBM, usually gives the best accuracy on tabular data but needs more careful tuning and can overfit if you add too many trees.

Question 14

Explain gradient descent in simple words. What does the learning rate do?

Accepted Answer

Imagine you're on a hill in fog and want to reach the bottom. You feel which way the ground slopes and take a step downhill, again and again. Gradient descent does the same with the model's error: the gradient tells us which direction increases the error, so we move the weights a small step in the opposite direction. The learning rate is the step size. Too big and you overshoot and bounce around; too small and training is very slow. In practice we use mini batches and optimisers like Adam.

Question 15

How does an LLM generate text?

Accepted Answer

An LLM is a neural network, a transformer, trained on huge amounts of text to predict the next token, which is a word or piece of a word. When you send a prompt, it's split into tokens, the model calculates probabilities for every possible next token, picks one, adds it to the text, and repeats until it's done. Settings like temperature control how random that choice is. Because it predicts plausible text rather than looking up facts, it can confidently produce wrong answers, which we call hallucinations.

Question 16

What are embeddings, and what are they used for?

Accepted Answer

An embedding is a list of numbers that represents the meaning of a piece of text, an image or an item. Things with similar meaning get vectors that are close together. So "how do I reset my password" and "forgot login" end up near each other even though they share no words. We use embeddings for semantic search, RAG, recommendations, clustering and finding duplicates. They're stored in a vector database, which can quickly find the nearest vectors using cosine similarity.

Question 17

What's the difference between a list, tuple, set and dictionary in Python? When would you use each?

Accepted Answer

A list is an ordered, changeable collection; good for sequences you add to, like rows to process. A tuple is ordered but can't be changed; good for fixed records like coordinates, and it can be a dictionary key. A set holds unique items with very fast membership checks; good for removing duplicates or checking "have I seen this ID?". A dictionary maps keys to values with fast lookup; good for things like user ID to user details. Checking membership in a set or dict is O(1) on average, but O(n) in a list.

Question 18

Explain INNER JOIN vs LEFT JOIN with an example.

Accepted Answer

An INNER JOIN returns only rows that match in both tables. A LEFT JOIN returns every row from the left table, plus matches from the right; where there's no match, the right side columns are NULL. Example: customers and orders. An inner join gives only customers who have ordered. A left join from customers gives all customers, and those with NULL order IDs are the ones who never ordered, which is a common way to find inactive customers.

Question 19

What are window functions? Explain ROW_NUMBER, RANK and DENSE_RANK.

Accepted Answer

Window functions calculate something across related rows without collapsing them like GROUP BY does. For example, a running total or each employee's salary rank within their department, using OVER (PARTITION BY department ORDER BY salary DESC). With ties: ROW_NUMBER gives unique numbers 1, 2, 3 even for ties, so add a tie breaker column to make it deterministic. RANK gives ties the same number and then skips, like 1, 1, 3. DENSE_RANK gives ties the same number without gaps, like 1, 1, 2. LAG and LEAD let you compare a row with the previous or next one.

Question 20

Write a query to get the top 3 products by revenue in each category.

Accepted Answer

I'd rank products within each category and then filter. First, in a subquery or CTE, compute revenue per product and add DENSE_RANK() OVER (PARTITION BY category ORDER BY revenue DESC) AS rnk. Then select rows WHERE rnk <= 3. I'd say out loud how I handle ties: DENSE_RANK keeps all tied products, which could return more than 3 rows; if they want exactly 3, I'd use ROW_NUMBER with a tie breaker like product ID.

Question 21

What's the difference between correlation and causation? How would you prove causation?

Accepted Answer

Correlation means two things move together; causation means one actually causes the other. Ice cream sales and drownings are correlated because both rise in summer, not because ice cream causes drowning. In products, users who use a new feature might retain better simply because they were already more engaged. To show causation, the best tool is a randomised A/B test: randomly give the feature to half the users, so the only systematic difference is the feature, then compare outcomes.

Question 22

How do you approach feature engineering for a new problem?

Accepted Answer

I start from the domain: what would a human expert look at to make this decision? For predicting late loan payments, that's things like income to EMI ratio, past late payments and recent changes in spending. Then I clean and encode: handle missing values, one hot encode categories, scale numbers if the model needs it, and create time based features like "days since last purchase". I check every feature for leakage and measure value with validation scores and feature importance. Usually a few good domain features beat many random ones.

Question 23

How do you handle an imbalanced dataset, like 1% fraud?

Accepted Answer

First, choose the right metric: precision, recall, PR AUC, not accuracy. Then options: use class weights so mistakes on the rare class cost more; resample, either undersample the majority or oversample the minority (SMOTE), only on the training data; and tune the decision threshold to hit the recall or precision the business needs. Also make sure the validation split keeps the same ratio, using a stratified split. Often, getting more examples of the rare class or better features helps most.

Question 24

How do you deal with missing values?

Accepted Answer

First I ask why they're missing, because that changes the fix. If income is missing because people didn't want to share it, the fact that it's missing may itself be a signal. Options: drop rows if very few are missing; fill with the median or most common value; fill using a model or a group average; or add a "was missing" flag column alongside the filled value. Tree models like XGBoost can handle missing values natively. Whatever I do, I fit the filling logic on the training data only.

Question 25

How do you choose which model to use?

Accepted Answer

I start with a simple baseline, like logistic regression or even a rule, so I know what "good" means. For tabular data, I'd then try gradient boosting because it's usually strongest. For images, text or audio, I'd use pretrained deep learning models or an LLM. Then I consider constraints: latency, cost, how much data I have, and whether the result must be explainable, for example in lending. I pick the simplest model that meets the target on validation data, because it's easier to maintain.

Question 26

What is cross validation and when would you not use standard k fold?

Accepted Answer

K fold cross validation splits training data into k parts, trains on k minus 1 and validates on the remaining part, rotating through all parts. Averaging gives a more reliable estimate than a single split, which matters with small data. I wouldn't use plain k fold for time series, because it lets the model see the future; I'd use time based splits instead. And if one user has many rows, I'd use group k fold so the same user isn't in both training and validation.

Question 27

How would you implement logistic regression from scratch?

Accepted Answer

I'd explain the steps, then code them in NumPy. One, initialise weights and bias to zeros. Two, compute z = X·w + b and the prediction p = sigmoid(z). Three, the loss is binary cross entropy. Four, the gradient is simple: dw = Xᵀ(p − y) / n and db = mean(p − y). Five, update w and b by subtracting learning rate times the gradient, and loop for some epochs. I'd add a check that the loss goes down, clip probabilities to avoid log of zero, and test on a tiny dataset where I know the answer.

Question 28

You've trained a model. How do you get it into production?

Accepted Answer

First, save the model together with its preprocessing, as one pipeline, with a version number in a model registry. Then wrap it in an API, for example with FastAPI, with input validation. Package it in a Docker image and deploy it behind a load balancer, on Kubernetes or a managed service. Before full rollout, test it in shadow mode or as a canary on a small share of traffic. Then monitor latency, errors, input drift and the business metric, with an easy way to roll back to the previous version.

Question 29

What is training serving skew, and how do you prevent it?

Accepted Answer

It's when the data the model sees in production is computed differently from the data it was trained on, so the model quietly performs worse. For example, in training "average spend" was computed over 30 days in SQL, but the live service computes it over 7 days in Python. Prevention: use one shared feature pipeline or a feature store for both training and serving, log the features used at prediction time and compare them with training data, and add tests that run the same input through both paths.

Question 30

When would you use prompting, RAG or fine tuning?

Accepted Answer

I go in order of cost. First, prompting: clear instructions and examples solve many problems and take hours, not weeks. If the model needs knowledge it doesn't have, like our policies or recent data, I use RAG, because it's easy to update and gives citations. I'd fine tune when the problem is behaviour, not knowledge: a very specific output style or format, a narrow task done at huge volume where a small fine tuned model is cheaper and faster, or domain language the model handles poorly. Often the best system combines RAG with a lightly fine tuned model.

Question 31

How do you evaluate an LLM application?

Accepted Answer

I start by defining what good looks like for the task, then build an evaluation set of real examples, say 50 to 200 to start, covering common cases, edge cases and known failures. For each, I decide how to score: exact checks with code where possible, like valid JSON or the right category; for open answers, a rubric scored by humans or by an LLM judge that I've checked against human ratings. For RAG I measure retrieval and answer quality separately. I run this suite on every prompt or model change, like unit tests, and I also watch real usage: thumbs down, edits and escalations.

Question 32

Explain self attention. Why do we divide by the square root of d_k?

Accepted Answer

Each token creates three vectors: a query, "what am I looking for?", a key, "what do I contain?", and a value, "what do I pass on?". For each token, we compare its query with every key using a dot product, which gives relevance scores. Softmax turns the scores into weights, and the output is a weighted sum of the values. So each token builds a new representation by looking at the others. We divide by the square root of d_k because dot products grow with dimension; without scaling, softmax becomes very peaked and its gradients become tiny, which slows training.

Question 33

What is LoRA, and why is it popular?

Accepted Answer

LoRA, low rank adaptation, fine tunes a big model by freezing the original weights and learning a small update for some weight matrices, written as the product of two thin matrices, A and B. Because the rank is small, you train maybe less than 1% of the parameters, which needs far less GPU memory and storage. You can keep one base model and swap small LoRA adapters for different tasks. QLoRA goes further by keeping the base model in 4 bit precision, so you can fine tune large models on a single GPU.

Question 34

What are vanishing gradients, and what fixed them in modern networks?

Accepted Answer

In deep networks, gradients are multiplied through many layers during backpropagation. If those factors are below one, the gradient shrinks to nearly zero by the time it reaches early layers, so they stop learning. Fixes that made deep networks work: activations like ReLU instead of sigmoid; residual or skip connections, which give gradients a direct path back; normalisation layers like LayerNorm; careful weight initialisation; and optimisers like Adam. Transformers rely heavily on residual connections and LayerNorm.

Question 35

What's the difference between data drift and concept drift? How do you detect them without labels?

Accepted Answer

Data drift is when the inputs change, like a new customer segment or a sensor recalibrated, while the relationship to the outcome is the same. Concept drift is when the relationship itself changes, like fraudsters changing tactics, so the same inputs now mean something different. Without fresh labels, I monitor input feature distributions against a training baseline using statistics like PSI or KS tests, watch the distribution of predictions and confidence scores, and track proxy business signals such as complaints or manual overrides. When labels arrive later, I compute real performance by time window.

Question 36

What would you monitor for a model in production?

Accepted Answer

Four layers. System health: latency, error rates, throughput, CPU, GPU and memory. Data: missing values, schema changes, feature distributions compared with training. Model: prediction distribution, confidence, and real accuracy when labels arrive. Business: the metric the model exists for, like fraud losses or conversion. Each alert should map to an action, like "page on call" or "retrain", otherwise people ignore alerts. I'd use something like Prometheus and Grafana for system metrics and a drift tool like Evidently for data.

Question 37

Click through rate dropped 20% overnight, but input distributions look normal. Model drift or a bug?

Accepted Answer

A sudden overnight drop smells like a bug or a change, not slow drift. I'd check, in order: did anything deploy, like a new model, feature pipeline or front end change? Is logging or the metric itself broken, for example clicks not being recorded? Are features arriving late or as defaults, which might not show in aggregate distributions? Is it all users or one platform, region or app version? I'd compare with a holdout or previous model if available. Only if everything checks out would I consider concept drift, like a holiday or competitor event, and retrain.

Question 38

How would you containerise a model and serve it on Kubernetes?

Accepted Answer

Write a small API, for example FastAPI, that loads the model once at startup and exposes a predict endpoint and a health check. Write a Dockerfile with a slim base image, pinned dependencies and the model either baked in or downloaded from a registry at start. On Kubernetes, create a Deployment with several replicas and resource requests and limits, readiness and liveness probes on the health endpoint, a Service in front, and a Horizontal Pod Autoscaler based on CPU or request rate. For GPUs, request nvidia.com/gpu resources and use GPU node pools.

Question 39

What does CI/CD look like for machine learning?

Accepted Answer

Like normal CI/CD, plus data and models. CI: on every code change, run unit tests, data validation tests and a quick training run on a sample to catch errors. Training pipelines run on a schedule or when new data arrives, log parameters and metrics to an experiment tracker, and register the model in a registry. CD: a new model is promoted only if it beats the current one on a fixed evaluation set and passes checks like bias and latency, then it's rolled out gradually with canary or shadow deployment, with automatic rollback if metrics drop.

Question 40

How do you make model training reproducible?

Accepted Answer

Version everything that affects the result: the code with git, the data with a snapshot or tool like DVC, the environment with a Docker image and pinned packages, and the configuration and hyperparameters in a tracked config file. Set random seeds. Log every run, with its code commit, data version, parameters, metrics and the resulting model, in an experiment tracker like MLflow. Then anyone can rerun or explain a model months later, which also matters for audits.

Question 41

What steps do you follow in a machine learning project, from start to finish?

Accepted Answer

One, understand the problem and the success metric with the people who need it. Two, collect the data and explore it: size, missing values, distributions and obvious errors. Three, clean the data and create features. Four, split into training, validation and test sets. Five, train a simple baseline, then better models, comparing them on validation data. Six, evaluate the final model once on the test set and check its mistakes. Seven, deploy it, for example as an API or a Streamlit app. Eight, monitor it and retrain when data changes.

Question 42

You get a new dataset. How do you explore it?

Accepted Answer

I start with the shape and column types, then look at the first few rows to understand what each column means. Next, missing values per column, duplicates, and basic statistics like mean, median, minimum and maximum, which reveal impossible values like negative ages. I plot distributions with histograms and count the categories. Then I look at relationships: how each feature relates to the target, and correlations between features. I write down surprises and questions for the data owner, because many problems are understood only by asking.

Question 43

How would you build a spam email classifier?

Accepted Answer

Clarify: what counts as spam, and which mistake is worse? Usually marking a real email as spam is worse, so precision matters. Data: emails labelled spam or not. Features: start with text turned into numbers, using TF IDF, plus simple signals like number of links or the sender's history. Model: a baseline with logistic regression or naive Bayes, which works well for text, then compare with a fine tuned small language model if needed. Evaluate with precision and recall on a test set, and pick a threshold that keeps false positives very low. Deploy it as a service, collect "not spam" corrections from users, and retrain regularly because spammers change tactics.

Question 44

Design a recommendation system for an e commerce site with 2 million products.

Accepted Answer

Clarify: where recommendations appear, the goal (clicks, purchases), latency budget and data available. Use two stages. Candidate generation: quickly narrow 2 million products to a few hundred, using collaborative filtering or two tower embeddings plus simple sources like popular in category. Ranking: a more detailed model, like gradient boosting or a neural ranker, scores candidates using user, product and context features, then business rules handle stock and diversity. Cold start: new users get popular and contextual items; new products use content features like category and text embeddings. Evaluate offline with recall and NDCG, then online with an A/B test on purchases. Serve with precomputed embeddings and a feature store for fresh features.

Question 45

Design a real time fraud detection system with a 150 ms latency budget.

Accepted Answer

Clarify the cost of missed fraud versus false declines, and the review team's capacity. Features: real time features like the number of transactions from this card in the last 10 minutes, computed from a stream, plus historical features from a feature store; accept that some features are slightly stale to meet latency. Model: gradient boosting, fast and strong on tabular data, with a threshold chosen for the business trade off, and a middle band sent to manual review. Serving: a low latency service with features cached in memory or Redis, and a fallback rule if the model is slow. Labels arrive late through chargebacks, so monitor drift and retrain regularly, and keep rules for brand new fraud patterns.

Question 46

Walk me through a project you're proud of.

Accepted Answer

Use STAR and go deep on your own decisions. "Situation: our support team was overwhelmed with repeated questions. Task: I wanted to build an assistant that answers from our help docs. Action: I built a RAG pipeline, but the first version answered only 60% of test questions correctly. I created an evaluation set of 80 real questions, found that most failures came from chunks splitting tables, switched to structure aware chunking and added reranking. Result: correct answers rose to 86%, and in a pilot it resolved about a quarter of tickets without an agent." Be ready for "why did you choose that?" at every step.

Question 47

Tell me about a time something you built failed.

Accepted Answer

Pick a real failure with a lesson, not a fake one. "I shipped a classification model that looked great offline, with 92% accuracy, but within two weeks users complained. I had split data randomly, and the same customers appeared in both training and test, so the score was too optimistic. I owned it with my manager, re split by customer, got a realistic 81%, and improved features to reach 86%. Since then, I always ask how the data is grouped before splitting, and I add a review of the evaluation setup to my checklist."

Question 48

Tell me about a disagreement with a teammate or stakeholder.

Accepted Answer

Show you disagree with data and respect. "A product manager wanted to launch our AI reply feature to all customers. I was worried because our tests showed it sometimes made up refund policies. Instead of just saying no, I showed five real examples and the error rate from our eval set, and proposed launching to 10% of tickets with agent approval first. We agreed on criteria to expand. Two weeks later the error rate was below our threshold and we expanded. We kept a good relationship because we focused on the shared goal."

Question 49

Explain a technical concept to me as if I'm not technical.

Accepted Answer

Pick something from your work and use an everyday analogy. For example, explaining RAG: "Imagine a very smart new employee who has read a lot but knows nothing about our company. If you ask them a policy question, they might confidently guess. RAG is like giving them the right pages of our policy manual before they answer, and asking them to point to the page. So answers are based on our actual documents, and you can check them." Then check understanding: "Does that make sense?"

Question 50

How do you keep up with the fast changes in AI?

Accepted Answer

Be specific and show you filter noise. "I follow a few high quality sources instead of everything: release notes from the main model providers, a couple of newsletters, and papers when something important comes up. But mostly I learn by building: when something new appears, like a new tool calling feature, I try it on a small project within a week and write a short note about what worked. I also review what changed in our own evals when we upgrade models, which teaches me more than reading headlines."

Question 51

How do you use AI coding assistants in your work, and how do you make sure the code is right?

Accepted Answer

Be honest and responsible. "I use AI assistants daily for boilerplate, tests, unfamiliar libraries and explaining code. But I treat the output like code from a new colleague: I read every line before accepting it, I make sure there are tests, I check security sensitive parts like authentication and SQL myself, and I never paste secrets or confidential data into tools that aren't approved. It makes me faster, but I'm still responsible for what I ship."

Question 52

Do you have any questions for us?

Accepted Answer

Always say yes and ask two or three real questions. Good ones: "What would success look like for this role in the first 6 months?", "What's the biggest challenge the team is facing with AI right now?", "How do you evaluate model or feature quality before shipping?", "How is the team split between building new features and maintaining existing ones?", and "What do people who do well here have in common?". Avoid asking about leave or salary in a technical round, and don't ask things that are on their website.

Question 53

What are your salary expectations?

Accepted Answer

Research first, using sites like Glassdoor, AmbitionBox, levels.fyi and people in similar roles, then give a range based on the role, not on your current pay. "Based on my research for this role and level in Bangalore, and my experience, I'm looking at a range of ₹22 to ₹28 lakh fixed. I'm flexible depending on the overall package, and I'm most interested in the role itself." If they insist on your current salary, you can share it, but steer back to the market range for the new role. Never accept or reject on the spot; ask for the offer in writing.

Question 54

What does len({'a': 1, 'b': 2, 'a': 3}) return?

Accepted Answer

2. Dictionary keys are unique. The second 'a' overwrites the first, leaving 2 keys.

Question 55

You need to check whether 1 million IDs are in a collection of 1 million known IDs. What's the efficient approach?

Accepted Answer

Put the known IDs in a set and use `in`. Membership checks in a set are O(1) on average, but O(n) in a list. With a list this becomes about a trillion comparisons.

Question 56

What does df.groupby('city')['amount'].mean() return?

Accepted Answer

One average amount per city. groupby splits the rows by city, then mean() averages 'amount' within each group, giving a Series indexed by city.

Question 57

Salaries in a team are mostly around ₹8 LPA, with one executive at ₹2 crore. Which best describes a typical salary?

Accepted Answer

The median. One extreme value pulls the mean up a lot. The median, the middle value, isn't affected by outliers.

Question 58

Users who turned on a new 'focus mode' spend 30% more time in the app. Can you say focus mode caused it?

Accepted Answer

No. People who opt in may already be more engaged, so run a randomised A/B test. This is self-selection: the opt-in group differs from everyone else. Only random assignment isolates the feature's effect.

Question 59

What does a 95% confidence interval for a model's accuracy mean?

Accepted Answer

If we repeated the evaluation on many fresh samples, about 95% of the intervals built this way would contain the true accuracy. A confidence interval describes the procedure's long-run coverage of the true value, not a per-prediction probability.

Question 60

You want only the cities with more than 100 orders. Which clause filters those groups?

Accepted Answer

HAVING COUNT(*) > 100. WHERE filters rows before grouping. HAVING filters groups after aggregation.

Question 61

How do you find customers who have never placed an order?

Accepted Answer

LEFT JOIN orders and keep rows where the order id IS NULL. A LEFT JOIN keeps every customer. Those with no match have NULL order columns.

Question 62

Joining 1M orders to a customers table returns 3M rows. What's the most likely cause?

Accepted Answer

Duplicate customer IDs in the customers table, so each order matches several rows. A join multiplies rows when the key isn't unique on one side. Check uniqueness before you join.

Question 63

Training accuracy is 98% and test accuracy is 65%. What's happening, and what's a reasonable fix?

Accepted Answer

Overfitting, so simplify the model, regularise or get more data. A large train/test gap means the model memorised the training data. Reduce its complexity, regularise, or add data.

Question 64

A fraud model is 99% accurate, and 1% of transactions are fraud. What's your first reaction?

Accepted Answer

Compare against the baseline and check precision and recall, because predicting 'not fraud' for everything also scores 99%. With rare classes, accuracy is misleading. Precision and recall show whether fraud is actually being caught.

ML Engineer interview questions and answers

How a ML Engineer interview usually goes

👋 Opening: about you

🧱 Fundamentals

🛠️ Role skills

🏗️ Design and case questions

🤝 Projects and behaviour

❓ Closing: your questions and salary

🧪 Quick quiz: test yourself

Python & coding

Statistics & data

SQL & data modelling

ML fundamentals

Serving & MLOps

Deep learning

Data engineering

Communication

ML system design

More questions to prepare

Practise with a score and gap analysis

Also for ML Engineer

Sources