Went through the full loop recently, sharing notes since I found these helpful when prepping.
Round 1 – coding + ads basics
- quick project deep dive + a few behavioral questions
- spent a good chunk walking through ads ecosystem end-to-end
- coding was a map-of-maps type problem, ended up using graph traversal + discussed optimizations
Round 2 – ads design + search + ML
- high-level ads system (think Bing Ads-ish)
- campaign management, pacing, ranking/quality score, auctions
- SSP integration + full ad serving flow
- search API design (“plumbers near X” type query)
- ingestion + indexing (inverted index + geo stuff)
- semantic matching + hybrid retrieval (BM25 + embeddings)
- ranking + personalization signals
- ML/LLM discussion
- random forest vs SVM
- RAG setups with LLMs
- tradeoffs between traditional ML vs LLM-heavy systems
- some talk around agent-style workflows for ads optimization
Round 3 – payment system design (razorpay-like)
- full payment flow (app → gateway → bank, OTP/3DS etc.)
- reliability stuff: idempotency, retries, handling timeouts, double charge prevention
- system concerns: fraud, rate limiting, ddos, observability
- integration + UX (localization, currency, etc.)
Round 4 – onsite (HM + bar raiser)
system design: analytics platform
- track user actions with ~18 month retention, 5-min granularity
- filtering by geo/language/etc
topics that came up:
- event schema (fact/dimension), partitioning, pruning
- sql vs time-series db
- pipelines + SLAs
- approximate structures (bloom filters, count-min sketch), redis hashing
- also drifted into some ML/LLM basics (transformers, sampling params, etc.)
Round 5 – onsite (LLD + behavioral)
- whiteboard LRU cache
- talked through eviction strategies + caching tradeoffs
- behavioral: why leaving, leadership/conflicts, notice period, etc.
overall: pretty system-design heavy loop, especially ads + data infra angles. coding wasn’t crazy hard but you still need to be clean.
Best Resources I used
LC Problem Selection: Grind 75
LC Solutions: different youtuber explaination videos helped a lot
System Design Strategy: Hello Interview: System Design Walkthroughs
Question bank for fresh non-leetcode questions: Offerretriever
happy to answer questions if anyone’s prepping