Topics Practice Study

Zenaique

Personalized AI interview prep. Structured questions across every GenAI sub-discipline. Practice, score, master.

Practice

Topics
Practice
Challenges
Dashboard

Browse

Formats
Roles
Companies
Top questions

Product

Pricing
Sign in

© 2026 Zenaique

Home/Roles/AI Researcher

AI Researcher

621 questions

Works on the science, papers, theory, training dynamics, novel architectures, scaling laws.

Sort

Questions

PremiumSpot the bug in…
Spot the ErrorMedium
Estimate the parameters weight tying saves a 128k-vocab model at d_model 2048
Multiple ChoiceEasy
Walk through what quadrupling vocab to 128k does to params, compute, and sequences
Short AnswerMedium
Complete the path from token id to the vector that enters block 1
Fill in BlankEasy
Derive the SwiGLU hidden width that matches a 4x GELU FFN parameter budget
Predict OutputMedium
PremiumPinpoint the mechanism that…
Multiple ChoiceMedium
Rescue a streaming deployment that collapses once early tokens leave the cache
Short AnswerHard
Reason about KV cache growth in a sliding window model pushed to 32k tokens
Multiple ChoiceMedium
Architect a layer pattern mixing sliding window and global attention for huge contexts
Short AnswerHard
Explain why raising the RoPE base theta helps a model accept longer contexts
Multiple ChoiceMedium
Pick the real reason Llama class models swapped LayerNorm for RMSNorm
Multiple ChoiceEasy
PremiumExplain how a layer…
Multiple ChoiceMedium
Spot the error in this transformer block that silently dropped its residuals
Spot the ErrorEasy
Define QK-norm and the training failure it prevents
FlashcardMedium
Diagnose attention logit explosion in a large run and defend QK-norm as the fix
Short AnswerHard
Spot the error in this description of pre-norm placement and the final norm
Spot the ErrorMedium
Diagnose why a 48-layer model diverges right after a pre-norm to post-norm refactor
Short AnswerMedium
Choose the positional scheme that fails hardest when context quadruples at inference
Multiple ChoiceMedium
Identify what changes when a block computes attention and FFN in parallel, PaLM style
Multiple ChoiceMedium
PremiumDebug a 64-layer stack…
Short AnswerMedium
Interpret what the logit lens reveals when you unembed intermediate layers
Multiple ChoiceHard
PremiumSelect every architecture choice…
Multi-selectMedium
Calculate the per token KV cache footprint of a Llama-3-8B style config in fp16
Predict OutputHard
Complete the per token KV cache memory formula
Fill in BlankMedium
Identify which tensors your inference server actually stores in the KV cache
Multiple ChoiceEasy

Showing 1–25 of 621

← PrevNext →

Cerebras Jpmorgan

Evenup Goldman Sachs

Datarobot Reliance Jio

Krutrim Mphasis

Persistent Polyai

Baseten Jump Trading

Ai4bharat Neptune Ai

Induced Ai Perplexity

Anyscale Goldman Sachs

Banana Dev Canva

Coreweave Harvey

Descript Jpmorgan

Anyscale Coreweave

Samsung Shield Ai

Fractal Analytics Mercor

Jane Street Robust Intelligence

Palantir Patronus

Ltimindtree Modal Labs

Mercor Mistral AI

Doordash Mistral AI

Coreweave Tencent

Fractal Analytics Hebbia