HclTech

Role: Data Scientist

College: Not Specified

Interview Date: February 1, 2025

Interview Type: Off-Campus (Applied via Naukri.com), Online (MS Teams)

Result: Cleared Round 1

Difficulty: Not Specified

Rounds: 1 (Technical Interview)

Topics Asked: Machine Learning, Dimensionality Reduction (multicollinearity, feature selection, PCA), Time Series Forecasting (non-stationarity, detection/removal, AR/MA parameters), Recommendation Systems, Naive Bayes (numerical vs. categorical features), Hyperparameter Tuning (GridSearchCV), GenAI/LLMs (OpenAI library authentication, prompting skills), Coding (row-wise sum of digits in CSV without standard library)

Experience:

The candidate applied for the Data Scientist position at HCL Tech through Naukri.com. The first round interview was conducted online via MS Teams.

Round 1: Technical Interview
The interview began with ID proof verification and a screenshot capture. This was followed by a basic introduction covering education and work history. The candidate then explained a machine learning project they had worked on.

The technical discussion included:

• Methods to reduce dimensionality when having 4000 columns/features and 1 target variable (discrete or continuous). The candidate explained multicollinearity elimination, feature selection, and PCA.
• Discussion about known machine learning models, specifically Time Series Forecasting and Recommendation Systems.
• In-depth questions about Time Series Forecasting, including non-stationarity, how to detect and remove it, and how to find values of AR(p), MA(q), and other parameters.
• Questions on supervised algorithms, particularly Naive Bayes, with scenarios involving numerical and categorical features.
• Discussion on hyperparameter tuning methods like GridSearchCV, its workings, and necessary considerations.

A real-time coding challenge was given where the candidate shared their screen and VSCode window.

• Coding Challenge: Given a comma-separated text file where the first row is the column title and remaining rows are numbers separated by commas. Perform a row-wise sum of all the digits without using any standard library.

Finally, questions were asked about experience in GenAI and LLMs, to which the candidate replied briefly about the authentication process for the OpenAI library and how to use it to train LLMs with good prompting skills. The candidate was advised to focus more on GenAI if selected for the next round, as the proposed position specifically demanded a majority of GenAI work.

The overall interview experience was described as great.