Home
Skills
ASC
McKinsey
Management Consultant
SWE
Google
Product Manager
SWE
Amazon
Logistics Network Strategy
IBA
Goldman Sachs
Post-MBA IB Analyst
PM
Microsoft
PM Interview
SWE
Stripe
Product Manager
SWE
Meta
Growth Analyst
UX
Apple
SWE Interview
Show 12 More Tracks…
Sign up free
ML Engineer
01
GPU Inference Bottleneck Diagnosis
GPU Inference Profiling
Diagnose and improve a GPU inference pipeline that misses its latency target
GPU INFERENCE PROFILING
Start
02
CUDA Parallelism Reasoning
CUDA Thread Mapping
03
Memory Bandwidth Optimization
Memory Coalescing
04
Tensor Operation Decomposition
Tensor Operation Mapping
05
Deep Learning Pipeline Architecture
Training Graph Design
06
Distributed Training Topology Design
Distributed Training Topology
07
Inference Latency Profiling
Layer Latency Profiling
08
Quantization Trade-off Analysis
Precision and Accuracy Trade-offs
09
Dynamic Batching Feasibility Assessment
Dynamic Batching
10
Transformer Attention Bottleneck Resolution
Attention Complexity
11
Production MLOps Orchestration
GPU Serving Allocation
12
NVLink Topology Exploitation
NVLink Topology
13
Sparse Model Acceleration Strategy
Sparsity Trade-offs
14
Custom Kernel Fusion Justification
Custom Kernel Fusion
15
Multi-GPU Scaling Bottleneck Diagnosis
Multi-GPU Scaling
16
LLM Token Throughput Maximization
Prefill and Decode Optimization
17
Compilation and Auto-tuning Trade-offs
JIT Compilation Cost
18
CUDA Library Selection
CUDA Library Selection
19
System-wide Performance Prediction
End-to-End Latency Modeling
20
ML Infrastructure Trade-offs
Cost-Latency-Accuracy Trade-offs
