314,552 interview questions from 6,000+ companies.
Tests learning agility under delivery pressure, with emphasis on ownership, prioritization, and adapting quickly to unfamiliar technical work.
Tests ownership during a production incident, including structured debugging, stakeholder communication, and learning from high-pressure technical problems.
Tests prioritization and decision-making under pressure, especially how you balance speed, quality, and long-term technical cost.
Design a real-time event pipeline that can handle millions of events per second with sub-second latency.
Tests understanding of ZeRO partitioning and memory savings across distributed training stages.
Tests GPU performance optimization skills for memory-bound CUDA kernels.
Tests your ability to reason about data layout effects on GPU caching and accelerator throughput.
Tests your strategy for improving reasoning in smaller models via training and optimization techniques.
Tests your ability to choose and balance distributed parallelism strategies for LLM training.
Tests your ability to diagnose and mitigate communication and scaling bottlenecks in distributed training.
Tests your understanding of long-context modeling changes for retrieval-focused transformer architectures.
Tests your ability to evaluate accuracy and performance impacts of quantization for code generation.
Tests your understanding of positional embeddings and extrapolation behavior for longer contexts.
Tests your understanding of GPU memory hierarchy and kernel optimization for matrix multiplication.
Tests your debugging approach for numerical issues in training pipelines and stability practices.
Tests system design for production-grade, low-latency streaming inference at Relace.
Tests knowledge of efficient attention implementations and their impact on memory traffic.
Tests your ability to design and implement faster inference using speculative decoding in production.
Tests your ability to build rigorous evaluation for code generation correctness beyond static metrics.
