Phase 9 · Platform & Operations
Search, observability, microservices, deployment and object storage.
- Search Systems & Elasticsearch: Inverted indexes, analyzers, BM25 relevance, autocomplete, fuzzy matching, CDC indexing pipelines, zero-downtime reindexing and cluster sizing.
- Observability: Logs, Metrics, Traces & SLOs: Structured logs, Prometheus metrics, OpenTelemetry traces, sampling, cardinality, SLOs, error budgets and burn-rate alerting, plus how to debug a 20-service system at 3am.
- Microservices Architecture & Monolith Migration: Monolith vs modular monolith vs microservices, DDD boundaries, data ownership, sync vs async, CQRS, event sourcing, strangler fig migration and the distributed-monolith trap
- Deployment, Kubernetes & Release Strategies: Images, probes, resources, autoscaling, canaries, feature flags, graceful shutdown and zero-downtime migrations — how code actually reaches production safely
- Object & File Storage: Block vs file vs object, S3 internals (data/metadata split, placement, erasure coding, checksums), multipart & presigned uploads, resumable uploads, dedup, lifecycle tiers, CDN delivery and video streaming.
- Key Technology Deep Dives: DynamoDB, Temporal & ZooKeeper: Three technologies interviewers love to probe once you name them: DynamoDB (keys, single-table design, GSI vs LSI, partition limits, transactions, DAX, Streams, global tables), Temporal durable execution (workflows vs activities, deterministic replay, versioning, history limits, when it beats saga + outbox), and ZooKeeper (ZAB, znodes, one-shot watches, recipes, and why Kafka and Kubernetes moved to KRaft and etcd).
- AI/LLM Fundamentals for Engineers: Transformers, Fine-tuning vs RAG vs LoRA, and Agent Protocols: The fundamentals underneath every LLM system design: tokens and BPE (and why Hindi or Tamil text costs more), embeddings and cosine similarity, the Transformer and self-attention worked through with real numbers, decoder-only generation and sampling, why long context is expensive, training vs fine-tuning vs inference, prompt engineering vs RAG vs fine-tuning, full fine-tuning vs LoRA vs QLoRA with parameter and memory arithmetic, evaluation basics, MCP vs A2A, and when an agent framework beats a plain loop.