Invideo AI Applied Scientist Interview: Rounds, Questions and Preparation

What the job involves

Applied Scientists in AI and deeptech build and deploy machine learning models that solve real-world problems, often working on generative AI, computer vision, or recommendation systems. At companies like Invideo AI, they collaborate with engineers and product teams to turn research ideas into scalable, production-ready features. The role blends coding, experimentation, and cross-functional communication.

Role and Technical Knowledge

1. 1

   Explain the difference between generative and discriminative models, and give an example of each relevant to video creation.

   Strong answers define both, contrast their objectives, and link to tasks like video generation (generative) versus video classification (discriminative).

2. 2

   How would you handle overfitting in a deep learning model for image or video generation?

   Cover techniques like data augmentation, dropout, regularization, early stopping, and using more diverse training data.

3. 3

   What are the key evaluation metrics for a text-to-video generation model, and why are they challenging?

   Mention FVD, IS, and human evaluation, and discuss the difficulty of quantifying creativity and temporal coherence.

4. 4

   Describe the architecture of a transformer and how it can be used in multimodal AI for video editing.

   Explain self-attention, encoder-decoder, and how transformers process text and visual tokens for tasks like video captioning or editing.

Practical Scenarios

1. 5

   You are given a dataset of user-uploaded videos with noisy labels for scene classification. How would you build a robust model?

   Discuss data cleaning, semi-supervised learning, noise-robust loss functions, and validation strategies.

2. 6

   A generative model you deployed produces artifacts in generated videos. How do you debug and improve it?

   Outline steps like inspecting training data, checking model capacity, adjusting loss functions, and using perceptual metrics.

3. 7

   How would you design an experiment to A/B test a new video recommendation algorithm?

   Define success metrics, control and treatment groups, sample size, and guard against biases like novelty effects.

4. 8

   Your model inference is too slow for real-time video editing. What optimizations would you apply?

   Cover model quantization, pruning, knowledge distillation, using ONNX or TensorRT, and batching or caching.

Behavioural and HR

1. 9

   Why do you want to work as an Applied Scientist in a generative AI startup like Invideo AI?

   Show passion for AI applications, mention specific products or challenges, and align with the company's mission.

2. 10

   Describe a time you worked in a team to solve a technical problem. What was your role?

   Use STAR method, highlight collaboration, your specific contribution, and the outcome.

3. 11

   How do you keep up with the latest research in AI and deeptech?

   Mention reading papers, following conferences like NeurIPS, online courses, and personal projects.

4. 12

   Tell me about a project where you had to learn a new technology quickly.

   Emphasize adaptability, resourcefulness, and the successful application of the new skill.

Coding and Algorithms

1. 13

   Implement a function to compute the intersection over union (IoU) of two bounding boxes in Python.

   Write clean code with edge case handling, and explain the formula and its use in object detection.

2. 14

   Given a list of video frames as numpy arrays, write code to compute the average frame difference.

   Show efficient numpy operations and discuss time complexity.

3. 15

   How would you implement a simple k-means clustering algorithm from scratch?

   Outline initialization, assignment, update steps, and convergence criteria, and mention its use in feature clustering.

4. 16

   Write a function to perform matrix multiplication without using built-in functions.

   Demonstrate understanding of loops and complexity, and mention optimization techniques like blocking.