Interview Experience – Senior Backend / Ads / Distributed Systems (5 Rounds)

---

🔹 Round 1 – Coding + Ads Fundamentals

Focus Areas:

• Project deep dive
• 3–4 behavioral questions
• Complete Ads ecosystem explanation
  • Advertiser → DSP → Ad Exchange → SSP → Publisher
  • Real-time bidding (RTB) flow
  • Ranking & auction basics
• End-to-end ad serving lifecycle

Coding Question:

• Map of Map encoding problem
• Required graph search
• Implemented, executed, and discussed optimizations

---

🔹 Round 2 – Ads HLD + Search + ML

1️⃣ Design HLD for Microsoft Bing Ads

Designed high-level architecture similar to Microsoft Bing Ads.

Covered:

• Campaign setup & management
• Budget pacing
• Ranking logic & quality score
• Auction mechanism
• Publisher integration via SSP
• End-to-end ads delivery flow

---

2️⃣ Search API Design

Problem:
Design an API for: “Plumbers near Kondapur”

Discussion Points:

• Data ingestion pipeline
• Indexing (inverted index + geo-spatial indexing)
• Semantic matching
• Hybrid retrieval (BM25 + embeddings)
• Ranking signals & personalization

---

3️⃣ ML & LLM Discussion

• Random Forest vs SVM
• RAG (Retrieval-Augmented Generation) with LLMs
• Trade-offs: classical ML vs LLM-based systems
• Agentic frameworks for automation in ads optimization

---

🔹 Round 3 – Payment System Design (Razorpay-like)

Designed a payment gateway system similar to Razorpay.

Covered:

Payment Flow

• App → Gateway → Bank
• OTP handling
• 3DS flow
• Callback/webhook handling

Reliability & Safety

• Idempotency keys
• Retry mechanisms
• Handling bank timeouts
• Double debit protection

System Concerns

• DDoS protection
• Fraud detection
• Rate limiting
• Observability

Integration & UX

• Integration across app, gateway, and bank APIs
• Localization & language support
• UI rendering based on user country/currency

---

🔹 Round 4 – Hiring Manager + Bar Raiser (Onsite)

System Design: Website Activity Analytics

Requirement:

• Track clicks/actions/users
• 18-month lookback
• 5-minute granularity
• Filter by language, country, and other attributes

Discussion Points:

• Event-based schema (fact + dimension tables)
• SQL vs Time-Series DB
• Partitioning strategy & pruning
• Logged-in vs logged-out user tracking
• Data pipelines & SLAs

Advanced Topics

• Bloom Filters
• Count-Min Sketch
• Hashing in Redis
• Sampling & standard deviation use cases
• Transformer architecture basics
• Effect of Temperature, Top-K, Top-P on LLM outputs
• AI tool adoption metrics & measuring agentic AI impact

---

🔹 Round 5 – (LLD) - Onsite

Coding on Whiteboard

• LRU Cache (write code on board)

System & Behavioral Discussion

• Eviction strategies
• Caching trade-offs
• Reason for leaving
• Leadership & conflict handling
• Notice period expectations