Interview Guide

Research Scientist Interview Guide - 2026 Edition

A structured preparation curriculum for Research Scientist roles at industrial AI labs, academic research groups, and applied ML organizations. Covers mathematical foundations through modern LLMs, deep RL, and trustworthy AI.

Author

Kundan Kumar

Published

2026-05-10

Preface

“Think mathematically, reason scientifically, and build efficiently.”

This is a structured preparation curriculum for Research Scientist roles at industrial AI labs, academic research groups, and applied ML organizations. It is built on a single conviction: every algorithm evolved to solve a pain point in its predecessor, and understanding that lineage is what separates a candidate who can recall methods from one who can reason about them.

The guide balances mathematical theory, statistical rigor, and hands-on implementation — tracing the evolution from classical machine learning through modern large language models and deep reinforcement learning systems.

The Lineage

Every method in this guide is positioned within an evolutionary chain:

Linear Models  →  Nonlinear Trees  →  Ensemble Boosting  →  Neural Networks
     →  Transformers  →  LLMs  →  DRL  →  Physics-Informed & Federated Systems

Each chapter answers: why was the next method needed, and what problem did it solve that its predecessor could not?

Learning Philosophy

  1. Start from intuition. Understand why each model exists before touching the math.
  2. Derive manually. Write gradients, losses, and activations by hand before running code.
  3. Implement from scratch. Rebuild models in PyTorch — no black boxes for core concepts.
  4. Visualize. Use loss landscapes, weight evolution plots, and activation maps to build spatial intuition.
  5. Communicate. Summarize each section as if presenting to a research committee.

Prerequisites

Readers are expected to have foundational knowledge in:

  • Mathematics — linear algebra, multivariate calculus, probability theory
  • Programming — Python, PyTorch or TensorFlow, basic data manipulation
  • Statistics — hypothesis testing, regression, inference
  • CS Fundamentals — data structures, algorithms, complexity analysis

Author

Kundan Kumar — https://kundan-kumarr.github.io/

Longer notes and weekly write-ups: kundan-kumarr.github.io/blog · neuravp.substack.com

Citation

Kumar, Kundan. (r substr(Sys.Date(), 1, 4)). From First Principles — Research Scientist Interview Guide. r substr(Sys.Date(), 1, 4) Edition.

License

This work is licensed under the MIT License.