System Design interviews are not limited to classic backend design. Depending on the role and specialization, companies conduct specialized sections for frontend, mobile, data and ML engineers. Each type has its own focus, patterns and evaluation criteria.
Prevalence of different types of interviews
Approximate estimates of how popular different types of System Design interviews are
Foundation
HTTP protocol
The base protocol for most backend services and APIs.
Backend System Design
Classic System Design
Focus on distributed systems, scalability and reliability of server infrastructure. This is the “gold standard” System Design interview.
Key Components
- Load Balancers
- SQL/NoSQL databases
- Caches (Redis, Memcached)
- Message Queues (Kafka, SQS)
- CDN, API Gateway
Typical tasks
Focus of assessment
Scalability, Availability, Consistency tradeoffs
NFR
Latency, Throughput, Fault Tolerance
Deep Dive
Sharding, Replication, CAP theorem
Frontend System Design
Client Application Architecture
Focus on the client ecosystem: component architecture, state management, rendering performance and UX. Framework used RADIO.
RADIO Framework
Key topics
- Component Architecture & Modularity
- State Management (Redux, Context, Zustand)
- SSR vs CSR vs SSG with Hydration
- Performance: LCP, FID, CLS
- Code Splitting & Lazy Loading
Typical tasks
Optimizations
Virtualization (windowing), Progressive Image Loading, Debouncing, Bundle optimization
Differences from Backend
Less focus on DB/infra, more on UX, accessibility, browser APIs
Foundation
Android: mobile OS
Mobile OS architecture and limitations for application design.
Mobile System Design
iOS/Android architecture
Central topics: working in unstable network conditions, battery and memory limitations, offline-first approach and data synchronization.
Offline-First
Local database as source of truth. Room/CoreData + Repository pattern
Battery & Memory
App Thinning, Image optimization, Background task management
Sync Strategies
Push Notifications, WebSockets, SSE, Conflict resolution
Architectural patterns
Typical tasks
Unique aspects of Mobile SD
Data Engineering System Design
Data Pipelines & Infrastructure
Focus on the data lifecycle: from sources through processing to storage and consumption. Batch vs Streaming, ETL vs ELT, Data Lakes vs Warehouses.
Data Journey - 6 layers
Sources
Processing
Storage
ML/AI
Consumers
Key Concepts
Spark/Airflow vs Kafka/Flink
Schema-on-write vs Schema-on-read
Debezium, real-time sync
Medallion Architecture
Immutable, append-only
Deduplicated, validated
Aggregated, ready for BI
Technologies
Typical tasks
ML System Design
MLOps & Model Serving
ML System Design combines the classic backend with ML-specific components: training pipelines, feature stores, model serving and drift monitoring.
ML Pipeline Architecture
Data Ingestion
Feature Store
Training
Model Registry
Serving
Feature Store
Centralized repository of features for training and inference. Solves the problem Training-Serving Skew.
For batch training
For real-time inference
Model Serving
Deployment Strategies
- Canary Rollout
- A/B Testing
- Shadow Deployment
- Blue-Green
Monitoring
- Data Drift detection
- Model Performance (P/R/AUC)
- Latency & Throughput
- Feature distribution
Typical tasks
Key question for ML SD interview
“How will you ensure consistency between the features used in training and inference?” → Feature Store with versioning and lineage tracking.
Comparison table
| Aspect | Backend | Frontend | Mobile | Data | ML |
|---|---|---|---|---|---|
| Main focus | Scalability | UX & Performance | Offline & Battery | Data Flow | Model Lifecycle |
| NFR priority | Latency, Throughput | LCP, FID, CLS | Battery, Memory | Freshness, Accuracy | Accuracy, Latency |
| State | DB, Cache | Redux, Context | Local DB + Sync | Warehouse/Lake | Feature Store |
| Framework | Alex Xu 4-step | RADIO | MVVM + Repository | 6-layer Data Journey | ML Pipeline |
| Typical duration | 45-60 min | 45 min | 45 min | 45-60 min | 45-60 min |
Key Findings
Backend SD - universal standard, but not the only format. Get ready for specialized sections.
Frontend SD focuses on client architecture, state management and web performance metrics.
Mobile SD - offline-first, sync strategies, and device limitations (battery, memory, network).
Data SD - batch vs streaming, ETL pipelines, and Medallion architecture for data lakes.
ML SD — Feature Stores, training-serving skew, model registry and drift monitoring.
Specify the format - ask before the interview recruiter, what type of System Design is expected for your role.