Entropik Technology Computer Vision Engineer Interview: Rounds, Questions and Preparation

What the job involves

A Computer Vision Engineer in AI and deeptech builds and deploys models that extract meaning from images and video, often for emotion recognition, human behavior analysis, or visual inspection. Day to day you will preprocess data, train deep learning models, optimize them for real-time performance, and integrate them into products or APIs. At fast-growing companies like Entropik Technology, you may also work on multimodal signals such as facial expressions, eye tracking, and gesture analysis.

Role and Technical Knowledge

1. 1

   What is the difference between object detection and image segmentation, and when would you use each in a product like emotion recognition?

   Explain that detection gives bounding boxes while segmentation gives pixel-level masks. For emotion recognition, segmentation can isolate the face or specific regions like eyes and mouth for finer analysis.

2. 2

   How does a convolutional neural network work? Walk me through the key layers and their purposes.

   Cover convolution, activation (ReLU), pooling, and fully connected layers. Mention how convolution extracts local features and pooling reduces spatial dimensions.

3. 3

   What are some common data augmentation techniques for image data, and why are they important?

   List flips, rotations, crops, color jitter, and noise. Explain that they increase dataset diversity, reduce overfitting, and improve generalization.

4. 4

   Explain the concept of transfer learning. How would you use a pre-trained model like ResNet for a new computer vision task?

   Describe using a model trained on ImageNet, freezing early layers, and fine-tuning later layers on your dataset. Mention that it saves data and compute.

Practical Scenarios and Problem Solving

1. 5

   You have a dataset of facial images but only 500 labeled samples. How would you build a robust emotion classifier?

   Suggest transfer learning, data augmentation, semi-supervised learning, or using a pre-trained face recognition model. Mention cross-validation and careful evaluation.

2. 6

   Your model performs well on the test set but poorly in real-time on a mobile device. What could be the reasons and how would you fix it?

   Consider overfitting, domain shift, latency, and model size. Propose quantization, pruning, using MobileNet, or optimizing inference with TensorRT or ONNX.

3. 7

   How would you handle class imbalance in a dataset for detecting rare facial expressions?

   Mention techniques like oversampling, undersampling, class weights, focal loss, and using metrics like F1-score or AUC instead of accuracy.

4. 8

   Describe how you would deploy a computer vision model as an API for a web application.

   Cover containerization (Docker), REST API with Flask or FastAPI, and cloud services. Mention load balancing, scaling, and monitoring.

Project and Experience Deep Dive

1. 9

   Tell me about a computer vision project you built. What was the problem, your approach, and the outcome?

   Use the STAR method. Highlight data collection, model choice, training details, evaluation metrics, and any deployment or impact.

2. 10

   In your project, what was the most challenging technical issue you faced and how did you resolve it?

   Pick a specific issue like overfitting, slow inference, or poor data quality. Explain your debugging process and the solution.

3. 11

   How did you evaluate your model's performance? Which metrics did you use and why?

   Discuss accuracy, precision, recall, F1, IoU for segmentation, or mAP for detection. Relate metric choice to the business goal.

4. 12

   If you had more time or resources, how would you improve that project?

   Suggest collecting more diverse data, trying newer architectures, hyperparameter tuning, or deploying and testing with real users.

Behavioural and HR Questions

1. 13

   Why do you want to work as a Computer Vision Engineer in a fast-growing AI company?

   Show passion for applied AI, mention interest in solving real-world problems, and connect to the company's domain like emotion AI or human behavior analysis.

2. 14

   Describe a time you had to learn a new technology quickly for a project or hackathon.

   Give a concrete example, such as learning PyTorch or OpenCV under a deadline. Emphasize resourcefulness and results.

3. 15

   How do you handle feedback or criticism on your work, especially when a model or approach is not performing well?

   Show openness to feedback, willingness to iterate, and a focus on data-driven improvement rather than personal attachment.

4. 16

   Where do you see yourself in three years, and how does this role fit into your career goals?

   Align your growth with the company's needs, such as deepening expertise in CV, leading projects, or contributing to product innovation.